JP2000041254A - Image decoding device and image decoding method - Google Patents

Abstract

(57) [Problem] To provide an MPEG down decoder which retains a base vector and eliminates a phase shift of pixels between a field DCT mode and a frame DCT mode. SOLUTION: When the DCT mode is the field mode, 4 × 4 reduced IDCT is performed. When the DCT mode is the frame mode, I DC is used for all coefficients of the DCT block.
DCT is performed to separate into two pixel blocks corresponding to interlaced scanning, and DCT is performed on each of the separated two pixel blocks. Then, the IDCT is performed on the low-frequency coefficients of the two pixel blocks, and the two pixel blocks are combined. When the DCT block is a base vector, IDCT is performed on the cosine curve of the base vector obtained as a result of performing IDCT on the coefficients of all frequency components, using the coefficients of the low-frequency components of the phase on the same curve.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to predictive coding by performing motion prediction in predetermined pixel blocks (macroblocks) and compression by performing orthogonal transformation in predetermined pixel blocks (orthogonal transformation blocks). The present invention relates to an image decoding apparatus and an image decoding method for decoding encoded first resolution compressed image data, and more particularly, to decoding a first resolution compressed image data and decoding a second resolution lower than the first resolution. The present invention relates to an image decoding device and an image decoding method for reducing the size of moving image data to a resolution of 1.

[0002]

2. Description of the Related Art MPEG2 (Moving Picture Experts G)
Standardization of digital television broadcasting using an image compression method such as roup phase 2) is in progress. Digital television broadcasting standards include standards corresponding to standard resolution images (eg, 576 effective lines in the vertical direction) and high resolution images (eg, 1152 effective lines in the vertical direction).
Book). Therefore, in recent years, by decoding compressed image data of a high-resolution image and reducing the compressed image data to half the resolution, image data of a standard-resolution image is generated, and the image data corresponds to the standard resolution. There is a demand for a down decoder for displaying on a television monitor.

[0003] M is obtained by performing predictive coding by motion prediction and compression coding by discrete cosine transform on a high-resolution image.
A down decoder that decodes a bit stream such as PEG2 and downsamples it to a standard resolution image is described in the document "Scalable Decoder without Low-Frequency Drift" (Iwahashi, Shinbayashi, and Kike: IEICE Tech.
8, 1995-01) (hereinafter, this document is referred to as Document 1). This document 1 discloses the following first to third down decoders.

The first down decoder, as shown in FIG. 48, calculates 8 (the number of coefficients counted from the DC component in the horizontal direction) .times.8 (counts from the DC component in the vertical direction) the bit stream of the high-resolution image. Inverse discrete cosine transform device 1001 for performing an inverse discrete cosine transform of the number of calculated coefficients, an adding device 1002 for adding a high-resolution image subjected to discrete cosine transform and a motion-compensated reference image, and a temporary reference image. A frame memory 1003 for storing and a frame memory 10
03, a motion compensator 1004 that performs motion compensation on the reference image stored at 03 with half-pixel accuracy, and a frame memory 1003.
And a down-sampling device 1005 for converting the reference image stored in to a standard resolution image.

In the first down decoder, an output image decoded as a high-resolution image by performing an inverse discrete cosine transform is reduced by a down-sampling device 1005 to output standard resolution image data.

The second down decoder, as shown in FIG. 49, performs a DCT (Disc) of a bit stream of a high-resolution image.
(Rete Cosine Transform) An inverse discrete cosine transform device 1011 for performing an 8 × 8 inverse discrete cosine transform by replacing coefficients of high frequency components of the block with 0, a high-resolution image subjected to discrete cosine transform and a reference image subjected to motion compensation , A frame memory 1013 for temporarily storing a reference image, a motion compensation device 1014 for performing motion compensation on the reference image stored in the frame memory 1013 with half-pixel accuracy, and a frame memory 1013 for storing And a downsampling device 1015 for converting the reference image into a standard resolution image.

In the second down-decoder, the down-sampler 1005 reduces the output image decoded as a high-resolution image by performing an inverse discrete cosine transform by replacing the high-frequency component coefficients among all the coefficients of the DCT block with 0. To output standard resolution image data.

As shown in FIG. 50, the third down decoder performs, for example, an inverse discrete cosine transform of 4 × 4 using only the coefficients of the low frequency component of the DCT block of the bit stream of the high resolution image to perform standard resolution. A reduced inverse discrete cosine transform device 1021 for decoding an image, an addition device 1022 for adding a standard resolution image subjected to reduced inverse discrete cosine transform and a motion-compensated reference image, and a frame memory 1023 for temporarily storing the reference image And the frame memory 10
23 is provided with a motion compensating device 1024 that performs motion compensation on the reference image stored therein with 1/4 pixel accuracy.

In the third down decoder, an inverse discrete cosine transform is performed using only low frequency component coefficients among all the coefficients of the DCT block, and a high resolution image is decoded as a standard resolution image.

Here, in the first down decoder,
Since the inverse discrete cosine transform is performed on all the coefficients in the DCT block to decode the high-resolution image, an inverse discrete cosine transform device 1001 having high arithmetic processing capability and a high-capacity frame memory 1003 are required. Also, in the second down decoder, since the high-frequency component of the coefficients in the DCT block is set to 0 to perform the discrete cosine transform and decode the high-resolution image, the inverse discrete cosine transform device 1
Although the arithmetic processing capacity of 011 may be low, a high-capacity frame memory 1013 is still required. In contrast to the first and second down decoders, the third down decoder performs an inverse discrete cosine transform using only low frequency component coefficients among all the coefficients in the DCT block. The calculation processing capacity of the memory 1021 may be low, and the capacity of the frame memory 1023 can be reduced because the reference image of the standard resolution image is decoded.

By the way, there are a progressive scanning method and an interlaced scanning method as a display method of a moving image such as a television broadcast. The progressive scanning method is a display method for sequentially displaying images obtained by sampling all pixels in a frame at the same timing. The interlaced scanning method is a display method in which images obtained by sampling pixels in a frame at different timings for each line in the horizontal direction are alternately displayed.

In this interlaced scanning method, one of images obtained by sampling pixels in a frame at different timings for each line is called a top field (also referred to as a first field), and the other is a bottom field (a second field). Field). An image including the first line in the horizontal direction of the frame is a top field, and an image including the second line in the horizontal direction of the frame is a bottom field. Therefore, in the interlaced scanning method, one frame is composed of two fields.

In MEPG2, in order to efficiently compress a moving image signal corresponding to the interlaced scanning method, not only a frame is assigned to a picture, which is a unit for compressing a screen, and encoding is performed, but also a field is assigned to the picture and encoded. You can also.

In MPEG2, when a field is assigned to a picture, the structure of the bit stream is called a field structure. When a frame is assigned to a picture, the structure of the bit stream is called a frame structure. In the field structure, a DCT block is formed from the pixels in the field, and the discrete cosine transform is performed for each field. This processing mode for performing the discrete cosine transform on a field basis is called a field DCT mode. In the frame structure, a DCT block is formed from pixels in a frame, and discrete cosine transform is performed on a frame basis. The processing mode in which discrete cosine transform is performed in frame units is referred to as frame D
Called CT mode. Furthermore, in the field structure, a macroblock is formed from the pixels in the field, and motion prediction is performed on a field basis. This processing mode for performing motion prediction in units of fields is called a field motion prediction mode. In the frame structure, a macroblock is formed from pixels in a frame, and motion prediction is performed in frame units. A processing mode in which motion prediction is performed in frame units is called a frame motion prediction mode.

[0015]

The above-mentioned reference 1
An image decoding apparatus that decodes compressed image data corresponding to the interlaced scanning method using the third down decoder described in “A Compensation Method of Drift”
Errors in Scalability ”(N. OBIKANE, K. TAHARAand JY
ONEMITSU, HDTV Work Shop '93) (this document is hereinafter referred to as Document 2).

In the conventional image decoding apparatus disclosed in Document 2, as shown in FIG. 51, a bit stream obtained by compressing a high-resolution image by MPEG2 is supplied, and a bit stream analyzing apparatus 1031 for analyzing the bit stream is provided.
A variable-length code decoding device 1032 for decoding a variable-length coded bit stream that allocates a code length according to the frequency of occurrence of data, and an inverse quantization device 1033 for multiplying each coefficient of the DCT block by a quantization step. , A reduced inverse discrete cosine transform device 1034 for decoding a standard resolution image by performing, for example, a 4 × 4 inverse discrete cosine transform using only coefficients of low frequency components among all coefficients of the DCT block.
And an adding device 103 for adding the reduced resolution discrete cosine transformed standard resolution image and the motion compensated reference image
5 and a frame memory 1036 for temporarily storing a reference image
And 1 / in the reference image stored in the frame memory 1036.
A motion compensator 1037 that performs motion compensation with 4-pixel accuracy.

The reduced inverse discrete cosine transform device 1034 of the conventional image decoding device shown in Document 2 performs inverse discrete cosine transform using only low frequency component coefficients among all the coefficients in the DCT block. , The positions of the coefficients for performing the inverse discrete cosine transform are different between the frame DCT mode and the field DCT mode.

Specifically, in the case of the field DCT mode, the reduced inverse discrete cosine transform device 1034
As shown in FIG. 2, of the 8 × 8 blocks in the DCT block,
The inverse discrete cosine transform is performed only on the low frequency 4 × 4 coefficients. In contrast, a reduced inverse discrete cosine transform device 1034
In the frame DCT mode, as shown in FIG. 53, among the 8 × 8 coefficients in the DCT block, 4 ×
Inverse discrete cosine transform is performed only on 2 + 4 × 2 coefficients.

The motion compensating device 1037 of the conventional image decoding device disclosed in Document 2 is based on information (motion vector) of motion prediction performed on a high-resolution image.
Motion compensation with 1/4 pixel accuracy corresponding to each of the field motion prediction mode and the frame motion prediction mode is performed. That is, although it is specified that motion compensation is performed with 1/2 pixel accuracy in MPEG2, the number of pixels in a picture is reduced to 1/2 when decoding a standard resolution image from a high resolution image. The motion compensator 10
At 37, the motion compensation is performed with the pixel accuracy of the motion compensation set to 1/4 pixel accuracy.

Therefore, the motion compensator 1037 performs linear interpolation on the pixels of the reference image stored in the frame memory 1036 as a standard resolution image to perform motion compensation corresponding to the high resolution image, and Pixels with pixel accuracy are generated.

More specifically, the linear interpolation of pixels in the vertical direction in the field motion prediction mode and the frame motion prediction mode will be described with reference to FIGS. In the drawings, the phase of a pixel in the vertical direction is shown in the vertical direction, and the phase at which each pixel of the display image is located is shown by an integer.

First, an interpolation process for an image for which motion prediction has been performed in the field motion prediction mode will be described with reference to FIG. For the high-resolution image (upper layer), as shown in FIG. 54 (a), motion compensation is performed with half-pixel accuracy independently for each field. In contrast,
For the standard resolution image (lower layer), FIG.
As shown in (b), linear interpolation is performed in the field based on the integer-precision pixels, and 1/4 pixels and 1/2
Pixels whose pixels are shifted in phase by 3/4 pixel are generated, and motion compensation is performed. That is, in the standard resolution image (lower layer), each pixel of 1/4 pixel precision of the top field is generated by linear interpolation based on each pixel of integer precision of the top field, and the bottom is generated based on each pixel of integer precision of the bottom field. Each pixel with 1/4 pixel precision of the field is generated by linear interpolation. For example, assume that the value of the pixel in the top field at the position where the vertical phase is 0 is a, and the value of the pixel of the top field at the position where the vertical phase is 1 is b. In this case, the pixel of the top field whose vertical phase is 1/4 is (3a + b) / 4.
And the top-field pixel whose vertical phase is 1/2 is (a + b) / 2, and the top-field pixel whose vertical phase is 3/4 is (a
+ 3b) / 4.

Next, an interpolation process of an image for which motion prediction has been performed in the frame motion prediction mode will be described with reference to FIG. For the high-resolution image (upper layer), as shown in FIG. 55A, interpolation processing is performed between fields, that is, interpolation processing is performed between the bottom field and the top field. Motion compensation is performed with pixel accuracy. For standard resolution images (lower layers)
As shown in FIG. 55 (b), the phases are shifted by 1/4 pixel, 1/2 pixel, and 3/4 pixel in the vertical direction based on the integer precision pixels of the two fields of the top field and the bottom field. The generated pixels are generated by linear interpolation, and motion compensation is performed. For example, the value of the pixel of the bottom field at the position where the vertical phase is −1 is a, the value of the pixel of the top field at the position where the vertical phase is 0 is b,
The value of the pixel of the bottom field whose vertical phase is 1 is c, the value of the pixel of the top field whose vertical phase is 2 is d, the bottom field whose vertical phase is 3 Let the value of the pixel of e be e. In this case, each pixel with a quarter-pixel accuracy whose vertical phase is between 0 and 2 is obtained as follows.

The pixel whose vertical phase is 1/4 is (a + 4b + 3c) / 8. Vertical phase is 1
The pixel at the position of / 2 is (a + 3c) / 4. A pixel having a vertical phase of 3/4 is (a + 2b + 3
c + 2d) / 8. A pixel having a vertical phase of 5/4 is (2b + 3c + 2d + e) / 8. The pixel whose vertical phase is 3/2 is (3c + e)
/ 4. A pixel having a vertical phase of 7/4 is (3c + 4d + e) / 8.

As described above, the conventional image decoding apparatus disclosed in the above-mentioned document 2 can decode compressed image data of a high-resolution image compatible with the interlaced scanning method into standard-resolution image data.

However, in the conventional image decoding apparatus disclosed in the above document 2, the phase of each pixel of the standard resolution image obtained in the field DCT mode is shifted from the phase of each pixel of the standard resolution obtained in the frame DCT mode. Specifically, in the field DCT mode, as shown in FIG. 56, the vertical phase of each pixel in the top field of the lower layer is 1/2, 5/2,. The vertical phase of the pixel is 1, 3
... On the other hand, in the frame DCT mode, as shown in FIG. 57, the vertical phase of each pixel in the top field of the lower layer is 0, 2,.
The vertical phase of each pixel in the bottom field of the lower layer is 1, 3,.... Therefore, images having different phases are mixed in the frame memory 1036, and the image quality of the output image is deteriorated.

Further, in the conventional image decoding apparatus disclosed in the above reference 2, the phase shift is not corrected between the field motion prediction mode and the frame motion prediction mode. Therefore, the image quality of the output image deteriorates.

The present invention has been made in view of the above circumstances, and eliminates a pixel phase shift between a field orthogonal transform mode and a frame orthogonal transform mode without impairing the interlacing property of an interlaced image. It is an object of the present invention to provide an image decoding device and an image decoding method for decoding standard resolution image data from compressed image data of a high resolution image that is capable of decoding.

[0029]

According to the present invention, there is provided an image decoding apparatus comprising: predictive coding by performing motion prediction on a predetermined pixel block (macroblock) basis; and a predetermined pixel block (orthogonal transform block) unit. An image decoding apparatus for decoding moving image data having a second resolution lower than the first resolution from compressed image data having a first resolution which has been compression-encoded by performing an orthogonal transformation in Inverse orthogonal transform means for performing inverse orthogonal transform on the coefficients of the low frequency components among the coefficients of the orthogonal transform block of the compressed image data, and compressed image data subjected to inverse orthogonal transform by the inverse orthogonal transform means. Adding means for adding the motion-compensated reference image data to output moving image data of the second resolution; and adding the moving image data output from the adding means to the reference image data. And a motion compensation unit for performing motion compensation on the macroblock of the reference image data stored in the storage unit. The inverse orthogonal transform unit includes: In the case of, for the curve obtained as a result of inverse orthogonal transformation of the coefficients of all frequency components of this orthogonal transformation block,
The inverse orthogonal transform is performed using the inverse orthogonal transform coefficient of the low-frequency component having the phase on the same curve.

In the image decoding apparatus, when the coefficients of the orthogonal transform block are base vectors, the curve obtained by inverse orthogonal transforming the coefficients of all the frequency components of the orthogonal transform block is on the same curve. Using the inverse orthogonal transform coefficient of the low frequency component of the phase, the inverse orthogonal transform is performed on the coefficient of the low frequency component among the coefficients of the orthogonal transform block. Then, the image decoding apparatus outputs moving image data having a second resolution lower than the first resolution and holding the basis vector components.

In the image decoding method according to the present invention, predictive coding is performed by performing motion prediction on a predetermined pixel block (macroblock) basis, and orthogonal transform is performed on a predetermined pixel block (orthogonal transform block) basis. From the compressed image data of the first resolution, which has been compression-encoded by
An image decoding method for decoding moving image data of a second resolution lower than the resolution of the compressed image data subjected to the orthogonal orthogonal transformation. Transform, add the compressed image data subjected to the inverse orthogonal transform, and the motion-compensated reference image data, store the obtained moving image data as reference image data, and store the stored reference image. When motion compensation is performed on a macroblock of data and the coefficients of the orthogonal transform block are base vectors, the same curve is obtained as a result of inverse orthogonal transform of the coefficients of all frequency components of the orthogonal transform block. The inverse orthogonal transform is performed using an inverse orthogonal transform coefficient of a low-frequency component having a phase on a curve.

In this image decoding method, when the coefficients of the orthogonal transform block are basis vectors, the same curve is obtained as the curve obtained as a result of inverse orthogonal transform of the coefficients of all frequency components of the orthogonal transform block. Using the inverse orthogonal transform coefficient of the low frequency component of the obtained phase, the inverse orthogonal transform is performed on the coefficient of the low frequency component among the coefficients of the orthogonal transform block. Then, in this image decoding method, moving image data having a second resolution lower than the first resolution and holding the base vector component is output.

[0033]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an image decoding apparatus to which the present invention is applied will be described as an embodiment of the present invention with reference to the drawings.

(First Embodiment) First, the first embodiment of the present invention will be described.
The image decoding apparatus according to the embodiment will be described.

As shown in FIG. 1, the image decoding apparatus 10 according to the first embodiment of the present invention receives a bit stream obtained by compressing a high-resolution image having 1152 effective lines in the vertical direction by MPEG2. Then, this device decodes the input bit stream and reduces it to half the resolution, and outputs a standard resolution image having 576 effective lines in the vertical direction, for example.

In the description of the embodiments of the present invention, a high-resolution image is also called an upper layer, and a standard-resolution image is also called a lower layer. Normally, when a DCT block having 8 × 8 discrete cosine coefficients is subjected to inverse discrete cosine transform, 8 × 8
Although it is possible to obtain decoded data composed of 8 pixels, for example, by decoding an 8 × 8 discrete cosine coefficient, 4 ×
The process of performing inverse discrete cosine transform and reducing the resolution to obtain decoded data composed of four pixels is called reduced inverse discrete cosine transform.

The image decoding apparatus 10 is supplied with a bit stream of a compressed high-resolution image, and analyzes the bit stream with a bit stream analyzing apparatus 11.
A variable-length code decoding device 12 that decodes the bit stream that has been subjected to the variable-length coding that assigns a code length according to the frequency of occurrence of data, and an inverse quantization device 13 that applies a quantization step to each coefficient of the DCT block. Have.

Further, the image decoding apparatus 10 performs a reduced inverse discrete cosine transform for a field mode to generate a standard resolution image by performing a reduced inverse discrete cosine transform on a DCT block which has been subjected to a discrete cosine transform in a field DCT mode and is not a base vector. A DCT that is a discrete cosine transform in the frame DCT mode and is not a base vector
A reduced inverse discrete cosine transform device 15 for frame mode that performs reduced inverse discrete cosine transform on a block to generate a standard resolution image, a standard resolution image subjected to reduced inverse discrete cosine transform, and a motion-compensated reference image. , A frame memory 17 for temporarily storing a reference image, a field mode motion compensator 18 for performing motion compensation corresponding to the field motion prediction mode on the reference image stored in the frame memory 17, and a frame memory. 1
7. A frame mode motion compensator 19 that performs motion compensation corresponding to the frame motion prediction mode on the reference image stored by
By performing post-filtering on the image stored in the frame memory 17, the image frame is converted, the phase shift of the pixel is corrected, and standard resolution image data to be displayed on a television monitor or the like is output. An image frame conversion / phase shift correction device 20 is provided.

The image decoding apparatus 10 further includes a basis vector discriminating apparatus 21 for discriminating whether or not the coefficients of the DCT block which have been subjected to the quantization step by the inverse quantization apparatus 13 are base vectors, and a field DCT mode. Is a discrete cosine transform and is a base vector DC
A field mode basis vector reduced inverse discrete cosine transform unit 22 for performing a reduced inverse discrete cosine transform on a T block to generate a standard resolution image, and a DC which is a basis vector subjected to a discrete cosine transform in a frame DCT mode.
And a frame mode basis vector reduced inverse discrete cosine transform unit 23 that performs reduced inverse discrete cosine transform on the T block to generate a standard resolution image.

The field mode reduced inverse discrete cosine transform unit 14 is used when a macroblock of an input bit stream is subjected to discrete cosine transform in the field DCT mode and a DCT block in the macroblock is not a base vector. The field mode reduced inverse discrete cosine transform unit 14 converts the 8 bits in the macroblock subjected to the discrete cosine transform in the field DCT mode.
For a DCT block showing × 8 coefficients, FIG.
2, inverse discrete cosine transform is performed only on the low-frequency 4 × 4 coefficients. That is, the reduced inverse discrete cosine transform is performed based on the discrete cosine coefficients of the four points in the horizontal and vertical low ranges. In the field mode reduced inverse discrete cosine transform device 14, by performing the reduced inverse discrete cosine transform as described above, one DCT block becomes 4 DCT blocks.
A standard resolution image composed of × 4 pixels can be decoded. As shown in FIG. 2, the phase of each pixel in the decoded image data is such that the vertical phase of each pixel in the top field is 1/2, 5/2,. The phases in the directions are 1, 3,.... That is, in the decoded top field of the lower layer, the phase of the first pixel (pixel having a phase of １ ／) is the first and second pixels (pixels of phase 0 and 2) from the top of the top field of the upper layer. And the phase of the second pixel from the top (pixel with a phase of 5/2) is the intermediate phase of the third and fourth pixels from the top of the top field of the upper layer (pixels with phases 4 and 6) Becomes In the bottom field of the decoded lower layer,
The phase of the first pixel (pixel with phase 1) is the intermediate phase between the first and second pixels (pixels with phases 1 and 3) from the top of the bottom field of the upper layer, and the second pixel from the top (phase is 1). The third and fourth pixels (phase 5) from the top of the bottom field of the upper layer
And 7 pixels).

The frame mode reduced inverse discrete cosine transform unit 15 is used when a macroblock of an input bit stream is subjected to discrete cosine transform in a frame DCT mode and a DCT block in the macroblock is not a base vector. The frame mode reduced inverse discrete cosine transform device 15 performs reduced inverse discrete cosine transform on a DCT block indicating 8 × 8 coefficients in a macroblock subjected to discrete cosine transform in the frame DCT mode. The reduced inverse discrete cosine transform unit 15 for frame mode decodes a resolution image in which one DCT block is composed of 4 × 4 pixels, and also outputs the standard resolution generated by the reduced inverse discrete cosine transform unit 14 for field mode. An image having the same phase as the phase of the pixel of the image is generated. In other words, as shown in FIG. 2, the vertical phase of each pixel of the top field is ２, 5/2, as shown in FIG. ...
The vertical phase of each pixel in the bottom field is 1, 3,
・ ・

The processing of the frame mode reduced inverse discrete cosine transform unit 15 will be described later in detail.

The field mode basis vector reduced inverse discrete cosine transform unit 22 is used when the macroblock of the input bit stream is subjected to discrete cosine transform in the field DCT mode and the DCT block in the macroblock is a basis vector. Can be Field mode basis vector reduced inverse discrete cosine transform device 22
Is a DCT showing 8 × 8 coefficients in a macroblock subjected to discrete cosine transform in the field DCT mode
The block is subjected to a two-dimensional matrix multiplication process in the horizontal and vertical directions using the inverse discrete cosine coefficient of the phase on the cosine curve obtained as a result of inverse discrete cosine transform of the coefficients of all frequency components of the DCT block. . This field mode basis vector reduced inverse discrete cosine transform device 2
In No. 2, by performing such a two-dimensional matrix operation, a standard resolution image in which one DCT block is composed of 4 × 4 pixels can be decoded. The phase of each pixel of the decoded image data is as shown in FIG.
The vertical phase of each pixel in the top field is 1/2,
.., And the vertical phase of each pixel in the bottom field is 1, 3,.

The field mode basis vector reduced inverse discrete cosine transform unit 22 processes 2
Details of the dimensional matrix will be described later.

The frame mode basis vector reduced inverse discrete cosine transform unit 23 is used when the macroblock of the input bit stream is subjected to discrete cosine transform in the frame DCT mode and the DCT block in the macroblock is a basis vector. Can be The frame mode basis vector reduced inverse discrete cosine transform unit 23 performs discrete cosine transform in a field DCT mode on a DCT block in which 8 × 8 coefficients in a macroblock subjected to discrete cosine transform in a frame DCT mode are indicated. Inverse discrete phase of the cosine curve obtained as a result of performing inverse discrete cosine transform of the coefficients of all the frequency components of the DCT block indicating the 8 × 8 coefficients in the transformed macroblock. Using the cosine coefficient, a two-dimensional matrix multiplication process in the horizontal and vertical directions is performed. The field mode basis vector reduced inverse discrete cosine transform device 22 can decode a standard resolution image in which one DCT block is composed of 4 × 4 pixels by performing such a two-dimensional matrix operation. it can. As shown in FIG. 2, the phase of each pixel of the decoded image data is 1 in the vertical direction of each pixel of the top field.
.., 5/2,..., And the vertical phase of each pixel in the bottom field becomes 1, 3,.

The details of the two-dimensional matrix processed by the frame mode basis vector reduced inverse discrete cosine transform unit 23 will be described later.

The basis vector discriminating device 21 decides whether or not the coefficient of the DCT block subjected to the quantization step by the inverse quantization device 13 is a basis vector. As a result of the discrimination by the basis vector discriminating device 21, D
If the coefficient of the CT block is not a base vector, the DC quantized by the inverse quantizer 13
The T block is supplied to the field mode reduced inverse discrete cosine transform device 14 or the frame mode reduced inverse discrete cosine transform device 15. If the coefficients of the DCT block are basis vectors, the DCT block that has been subjected to the quantization step by the inverse quantization unit 13 is converted into the field mode basis vector reduced inverse discrete cosine transform unit 22 or the frame mode basis vector. It is supplied to the reduced inverse discrete cosine transform unit 23.

The details of this determination will be described later.

The adder 16 includes a field mode reduced inverse discrete cosine transform device 14, a frame mode reduced inverse discrete cosine transform device 15, a field mode basis vector reduced inverse discrete cosine transform device 22, or a frame mode basis vector reduced inverse cosine transform device. If the macroblock subjected to the inverse inverse discrete cosine transform by the discrete cosine transform unit 23 is an intra image, the intra image is stored in the frame memory 17 as it is. Further, the adding device 16
Field mode reduced inverse discrete cosine transform device 14,
The macroblock subjected to the reduced inverse discrete cosine transform by the frame mode reduced inverse discrete cosine transform device 15, the field mode basis vector reduced inverse discrete cosine transform device 22, or the frame mode basis vector reduced inverse discrete cosine transform device 23 is an inter image. In some cases, the motion compensation device for field mode 1
8 or a reference image which has been motion-compensated by the frame mode motion compensator 19, and
To be stored.

The field mode motion compensator 18
This is used when the motion prediction mode of the macroblock is the field motion prediction mode. The field mode motion compensator 18 is configured to obtain a reference image of a standard resolution image stored in the frame memory 17 with 1/4 pixel accuracy in consideration of a phase shift component between a top field and a bottom field. Interpolation processing is performed to perform motion compensation corresponding to the field motion prediction mode. The reference image which has been motion-compensated by the field mode motion compensator 18 is supplied to the adder 16 and is synthesized with the inter image.

The frame mode motion compensator 19 is used when the macroblock motion prediction mode is the frame motion prediction mode. The frame-mode motion compensator 19 is configured to perform a ４ of a reference image of a standard resolution image stored in the frame memory 17 in consideration of a phase shift component between a top field and a bottom field.
Interpolation processing is performed with pixel accuracy, and motion compensation corresponding to the frame motion prediction mode is performed. The reference image which has been motion-compensated by the frame mode motion compensator 19 is supplied to the adder 16 and is synthesized with the inter image.

The image frame conversion / phase shift correction device 20 is supplied with a standard resolution reference image stored in the frame memory 17 or an image synthesized by the addition device 16, and this image is subjected to post-filtering to form a top field and a bottom field. Is corrected, and the image frame is converted so as to conform to the standard of the television having the standard resolution. That is, the image frame conversion / phase shift correction device 20
Is that the vertical phase of each pixel in the top field is 1 /
The standard resolution image in which the vertical phase of each pixel in the bottom field is 1, 3... And the vertical phase of each pixel in the top field is 0,
Are corrected so that the vertical phase of each pixel in the bottom field becomes 1, 3, 5,.
In addition, the image frame conversion / phase shift correction device 20 reduces the image frame of the high-resolution television standard to 1/4 and converts the image frame to the image frame of the standard resolution television standard.

The image decoding apparatus 10 according to the first embodiment of the present invention has the above-described configuration, thereby decoding a bit stream obtained by compressing a high-resolution image by MPEG2 and reducing the resolution to half. Thus, a standard resolution image can be output.

Next, the processing contents of the frame mode reduced inverse discrete cosine transform unit 15 will be described in more detail.

The frame-mode reduced inverse discrete cosine transform unit 15 can perform one or both of the following one-block processing and two-block processing. The frame mode reduced inverse discrete cosine transform device 15 may perform one block processing or two
The block processing may be switched and used, or only one of the processing may be performed.

First, the one-block processing will be described.
FIG. 3 shows a diagram for explaining the contents of one block processing.

As shown in FIG. 3, a bit stream obtained by compression-encoding a high-resolution image is input to the frame mode reduced inverse discrete cosine transform unit 15 in units of one DCT block.

First, at step S1, the discrete cosine coefficient y of this one DCT block (the vertical coefficient among all the discrete cosine coefficients of the DCT block is y ₁
As ~y ₈ shown in FIG. ) Is subjected to an 8 × 8 inverse discrete cosine transform (IDCT 8 × 8). By performing an inverse discrete cosine transform, 8 × 8 decoded pixel data x (vertical pixel data among all the pixel data of the DCT block is shown as x ₁ to x ₈ in the drawing) is obtained. Can be.

Subsequently, in step S2, this 8 ×
8 pixel data x are alternately taken out line by line in the vertical direction, and a 4 × 4 top-field pixel block corresponding to interlaced scanning and a 4 × 4 pixel block corresponding to interlaced scanning are extracted.
Are divided into two pixel blocks of the bottom field. That is, the pixel data x ₁ of the first line in the vertical direction, takes out the pixel data x ₃ of the third line, the pixel data x ₅ of the fifth line, the seventh line of the pixel data x _7, the top field Is generated. Further, the pixel data x ₂ of the second line in the vertical direction, takes out the pixel data x ₄ of the fourth line, the pixel data x ₆ of 6 line, the eighth line of the pixel data x _8, the bottom field Is generated. The process of separating each pixel of the DCT block into two pixel blocks corresponding to interlaced scanning is hereinafter referred to as field separation.

Subsequently, in step S3, each of the two pixel blocks subjected to field separation is 4 × 4
Is performed (DCT4 × 4).

Subsequently, in step S4, the discrete cosine coefficient z of the pixel block corresponding to the top field obtained by performing the 4 × 4 discrete cosine transform (the vertical cosine coefficient of all the coefficients of the pixel block corresponding to the top field) The high-frequency components of the discrete cosine coefficients in the directions are shown in the figure as z ₁ , z ₃ , z ₅ , and z _7. ) Are thinned out to obtain a pixel block composed of 2 × 2 discrete cosine coefficients. Also, 4x
4, the discrete cosine coefficient z of the pixel block corresponding to the bottom field obtained by performing the discrete cosine transformation of 4 (the vertical discrete cosine coefficients among all the coefficients of the pixel block corresponding to the bottom field are z ₂ , z ₄ , High-frequency components (shown in the figure as z ₆ and z ₈ ) are thinned out to obtain a pixel block composed of 2 × 2 discrete cosine coefficients.

Subsequently, in step S5, for the pixel block from which the discrete cosine coefficient of the high frequency component has been thinned,
Perform 2 × 2 inverse discrete cosine transform (IDCT2 × 2). By performing a 2 × 2 inverse discrete cosine transform, 2
'The vertical pixel data of all the pixel data of the (top field pixel block x' decoded pixel data x × 2 _1, x _'3 as shown in the figure, also, the pixels corresponding to the bottom field shown in FIG vertical pixel data of all the pixel data of block x as _'2, x' _4.) can be obtained.

Subsequently, in step S6, the pixel data of the pixel block corresponding to the top field and the pixel data of the pixel block corresponding to the bottom field are alternately combined line by line in the vertical direction to obtain 4 × 4 pixels. A DCT block that has been subjected to a reduced inverse discrete cosine transform composed of pixel data of The process of alternately combining the pixels of the two pixel blocks corresponding to the top field and the bottom field in the vertical direction is hereinafter referred to as frame combining.

By performing the one-block processing shown in steps S1 to S6, the reduced inverse discrete cosine transform unit for frame mode 15 can reduce the inverse discrete cosine transform device for field mode as shown in FIG. 1
4. It is possible to generate a 4 × 4 DCT block composed of pixels having the same phase as the pixel of the standard resolution image generated in step 4.

In the frame mode reduced inverse discrete cosine transform unit 15, one block process from step S1 to step S6 is performed using one matrix. Specifically, in the frame mode reduced inverse discrete cosine transform device 15, a matrix [F shown in the following equation 1 obtained by expanding and calculating the above processing using the addition theorem
S ′] and the discrete cosine coefficient y of one DCT block
By performing a matrix operation on (y _{1 to} y ₈ ), pixel data x ′ of the DCT block subjected to the reduced inverse discrete cosine transform is obtained.
(X ′ _{1 to} x ′ ₄ ) can be obtained.

[0066]

(Equation 1)

However, in the formula (1), A to J are as follows.

[0068]

(Equation 2)

Next, the two-block processing will be described. FIG. 4 shows a diagram for explaining the contents of the two-block processing.

As shown in FIG. 4, a bit stream obtained by compressing and encoding a high-resolution image is input to the frame mode reduced inverse discrete cosine transform unit 15 in units of two DCT blocks. For example, when the macro block has a so-called 420 format including a DCT block of four luminance components and a DCT block of two color difference components, a DCT block of two luminance components (Y) adjacent vertically is used. Is entered. When the macro block is configured as shown in FIG. 5, DCT block 0 and DCT block 2 of the luminance component (Y) are input as a pair, and DCT block 1 and DCT block 3 are input. Entered in pairs.

First, in step S11, two D
Discrete cosine coefficient y of the CT block (DC time before
Among the discrete cosine coefficients of the T block, the coefficients in the vertical direction are shown in the figure as y ₁ to y ₈ , and the DCT after time
The coefficients in the vertical direction among all the discrete cosine coefficients of the block are shown in the figure as y ₉ to y ₁₆ . ) For each of the 8 × 8 inverse discrete cosine transforms (IDCT8 ×
Perform 8). By performing the inverse discrete cosine transform, 8
× 8 decoded pixel data x (vertical pixel data among all the pixel data in the temporally previous DCT block is shown in the figure as x ₁ to x ₈ , and all the temporally subsequent DCT blocks are shown in the figure. Of the pixel data in the vertical direction are shown as x ₉ to x ₁₆ in the figure).

Subsequently, in step S12, the 8.times.8 pixel data x of the two DCT blocks are alternately extracted line by line in the vertical direction, and the 8.times.8 pixel data x of the top field corresponding to the interlaced scanning are extracted. The field is divided into two pixel blocks of an 8 × 8 pixel block of a bottom field corresponding to the interlaced scanning. That is, from the DCT block temporally before, the pixel data x ₁ of the first line and the pixel data x ₃ of the third line in the vertical direction
When a fifth line of the pixel data x _5, takes out the pixel data x ₇ 7 line, from the DCT block after time, the pixel data x ₉ the first line in the vertical direction, of the third line Pixel data x ₁₁ and pixel data x on the fifth line
_13, 7 takes out the pixel data x ₁₅ of the line, and generates a pixel block of 8 × 8 corresponding to the top field. Moreover, the temporally previous DCT block, the pixel data x ₂ of the second line in the vertical direction, 4 the pixel data x ₄ in line, the pixel data x ₆ in the sixth line, 8 pixels on Line Data Remove the and x _8, DCT after temporally
From the block, pixel data x ₁₀ of the second line in the vertical direction
When, a fourth line of pixel data x _12, the sixth line of the pixel data x _14, takes out the pixel data x ₁₆ of the eighth line, and generates a pixel block corresponding to the bottom field.

Subsequently, in step S13, an 8 × 8 discrete cosine transform (DCT 8 × 8) is performed on each of the two 8 × 8 pixel blocks separated into fields.

Subsequently, in step S14, 8 × 8
The discrete cosine coefficient z of the pixel block corresponding to the top field obtained by performing the discrete cosine transformation of (the vertical discrete cosine coefficient among all the coefficients of the pixel block corresponding to the top field is z ₁ , z ₃ , z _5, z _7,
These are shown in the figure as z ₉ , z ₁₁ , z ₁₃ , and z ₁₅ . ) Is thinned out to obtain a pixel block composed of 4 × 4 discrete cosine coefficients. Also, the discrete cosine coefficient z of the pixel block corresponding to the bottom field obtained by performing the 8 × 8 discrete cosine transform (the discrete cosine coefficient in the vertical direction among all the coefficients of the pixel block corresponding to the bottom field is z ₂ , Z ₄ , z ₆ , z ₈ , z ₁₀ , z ₁₂ , z ₁₄ , z
_It is shown in the figure as ₁₆ . ) Is thinned to form a pixel block composed of 4 × 4 discrete cosine coefficients.

Subsequently, in step S15, the 4 × 4 inverse discrete cosine transform (ID) is performed on each of the 4 × 4 pixel blocks obtained by thinning out the discrete cosine coefficients of the high frequency components.
CT4 × 4). By performing 4 × 4 inverse discrete cosine transform, 4 × 4 decoded pixel data x ′ (vertical pixel data of all pixel data of the pixel block corresponding to the top field is x ′ ₁ , x ' ₃ , x'
_5, x _'7 as shown in the figure, also, the vertical pixel data x of all the pixel data of the pixel block of the bottom field' figure as _2, x _'4, x' _6, x _'8 Shown inside. ) Can be obtained.

Subsequently, in step S16, the frame data of the pixel data of the pixel block corresponding to the top field and the pixel data of the pixel block corresponding to the bottom field are alternately framed line by line in the vertical direction.
A DCT block that has been subjected to a reduced inverse discrete cosine transform composed of 8 × 8 pixel data is generated.

By performing the two-block processing shown in steps S11 to S16, the reduced inverse discrete cosine transform unit for frame mode 15 is used in the reduced inverse discrete cosine transform unit for field mode as shown in FIG. A DCT block composed of pixels having the same phase as the phase of the pixel of the standard resolution image generated in 14 can be generated.

In the frame mode reduced inverse discrete cosine transform unit 15, the two block processes from step S11 to step S16 are calculated using one matrix. Specifically, in the frame mode reduced inverse discrete cosine transform device 15, the matrix [FS ″] shown in the following equation (2) obtained by expanding the above processing using the addition theorem, The matrix operation is performed on the discrete cosine coefficients y (y _{1 to} y ₁₆ ) of the two DCT blocks, and the reduced inverse discrete cosine transformed pixel data x ′ of the DCT block is obtained.
(X ′ _{1 to} x ′ ₈ ) can be obtained.

[0079]

(Equation 3)

However, in this equation (2), A to D are
It is as follows.

[0081]

(Equation 4)

[0082]

(Equation 5)

[0083]

(Equation 6)

[0084]

(Equation 7)

In this equation (2), a to g are
It is as follows.

[0086]

(Equation 8)

In the frame mode reduced inverse discrete cosine transform unit 15, the so-called 420 shown in FIG.
When a macro block of the format is input,
Steps S11 to S1 are performed for the luminance component.
The reduced inverse discrete cosine transform is performed by performing the two-block processing shown in FIG.
The reduced inverse discrete cosine transform may be performed by performing the one-block processing shown in step S6.

In the image decoding apparatus 10, the 4 × 4
And the above-described step S1 of the reduced inverse discrete cosine transform device 15 for frame mode.
The reduced inverse discrete cosine transform processing by one block processing in step S6 may be performed using a high-speed algorithm.

For example, Wang's algorithm (reference: Zhong DE Wang., "Fast Algorithms for the Discre
te W Transform and for the Discrete Fourier Transf
orm ", IEEE Tr. ASSP-32, NO. 4, pp. 803-816, Aug. 1984) can speed up the processing.

When the matrix operated by the field mode reduced inverse discrete cosine transform unit 14 is decomposed using the Wang algorithm, the matrix is decomposed as shown in the following equation (3).

[0091]

(Equation 9)

FIG. 6 shows a processing flow in the case where the Wang algorithm is applied to the processing of the field mode reduced inverse discrete cosine transform device 14. As shown in this processing flow, high speed can be realized by using the first to fifth multipliers 14a to 14e and the first to ninth adders 14f to 14n.

When the matrix [FS '] operated by the reduced inverse discrete cosine transform unit for frame mode 15 is decomposed using the Wang algorithm, it is decomposed as shown in the following equation (4).

[0094]

(Equation 10)

However, in this equation (4), A to J are
It is as follows.

[0096]

[Equation 11]

FIG. 7 shows a processing flow when the Wang algorithm is applied to the processing of the frame mode reduced inverse discrete cosine transform unit 15. As shown in this processing flow, high speed can be realized using the first to tenth multipliers 15a to 15j and the first to thirteenth adders 15k to 15w.

Next, the details of the determination of the base vector by the base vector determination device 21 will be described in more detail.

Here, the coefficients in the DCT block in which the 8 × 8 DCT coefficients dequantized by the dequantizer 13 are shown are represented by a coefficient group A, a coefficient group B, and a coefficient group as shown in FIG. Divide into C. Specifically, the coefficient group A is such that the horizontal spatial frequency or the vertical spatial frequency of the DCT coefficient is DC
(DC component) are 15 coefficients. The coefficient C is D
The horizontal spatial frequency and the vertical spatial frequency of the CT coefficient are the highest coefficient. The coefficient group B is the other remaining 48 coefficients.

As shown in FIG. 9, when all of the following conditions 1 to 3 are true, the basis vector discrimination device 21 decides that the DCT coefficient of this DCT block is a basis vector.

Condition 1: Of the 15 coefficients in coefficient group A, 14 are 0 and only 1 is not 0. Condition 2: All 48 coefficients of the coefficient group B are 0. Condition 3: the coefficient C is 0 or 1.

The reason why the coefficient C is determined as the condition 3 is to cope with a case where 1 is added when the sum of all the coefficients becomes an even number. Thus, the mismatch of the inverse discrete cosine transform is determined. Deterioration of image quality due to accumulation of errors can be prevented.

The basis vector determining device 21 determines whether the coefficients in the DCT block are the basis vectors as described above. If it is determined that the DCT block is not a base vector, the DCT block is subjected to the reduced inverse discrete cosine transform by the field mode reduced inverse discrete cosine transform unit 14 or the frame mode reduced inverse discrete cosine transform unit 15. When the DCT block is a basis vector, the DCT block is subjected to a reduced inverse discrete cosine transform by the field mode basis vector reduced inverse discrete cosine transform unit 22 or the frame mode basis vector reduced inverse discrete cosine transform unit 23.

Next, the two-dimensional matrix in which the DCT block of the basis vector is multiplied in the horizontal and vertical directions by the field mode basis vector reduced inverse discrete cosine transform unit 22 will be described in further detail.

FIGS. 10 to 16 show cosine curves based on first to seventh order basis vectors in the field mode, and show the phase relationship between ordinary 8-point inverse discrete cosine coefficients and reduced inverse discrete cosine coefficients for basis vectors. Will be described.
10 shows a cosine curve based on a primary basis vector, FIG. 11 shows a cosine curve based on a secondary basis vector, FIG. 12 shows a cosine curve based on a tertiary basis vector, and FIG. 14 shows a cosine curve based on a fifth-order basis vector, FIG. 15 shows a cosine curve based on a sixth-order basis vector, and FIG. 16 shows a cosine curve based on a seventh-order basis vector. The cosine curve is shown. Note that the horizontal axis of each figure shows the phase of the pixel after the inverse discrete cosine transform. In addition, a white circle in the figure indicates 8 in the normal processing in which no reduction is performed.
A point inverse discrete cosine coefficient is shown, and a black circle shows a reduced inverse discrete cosine coefficient for a base vector.

As shown by white circles in FIGS. 10 to 16, the eight-point inverse discrete cosine coefficient calculated by the normal processing without reduction is such that the phase on the cosine curve based on the first to seventh order basis vectors is an integer. (0-7). Therefore, the pixel in this DCT block after performing the inverse discrete cosine transform using the 8-point inverse discrete cosine coefficient has a base vector component.

On the other hand, the reduced inverse discrete cosine coefficient for the basis vector for performing the reduction processing is the same as the cosine curve on the cosine curve based on the base vector of the normal processing, as shown by the black circles in FIGS. This is the value at the position where the phase of the pixel is 0.5, 2.5, 4.5, 6.5. The field mode basis vector reduced inverse discrete cosine transform device 22 performs a matrix operation in the horizontal and vertical directions using the reduced inverse discrete cosine coefficient, thereby holding the basis vector components similar to those in the normal processing and performing the reduction. Inverse discrete cosine transform can be performed.

The matrix of the reduced inverse discrete cosine coefficient calculated by the field mode basis vector reduced inverse discrete cosine transform unit 22 will be specifically described below.

First, an ordinary matrix of 8-point inverse discrete cosine coefficients is shown in the following equation (5).

[0110]

(Equation 12)

The matrix of this equation (5) indicates the inverse discrete cosine coefficient of the phase angle π to 105π with the denominator being 16. In this matrix, the first column to the seventh column correspond to the order of the first to seventh base vectors, with the leftmost column as the zero column. The matrix of this equation (5) is converted to a phase angle of 0 to π /
When Cos (π / 4) and Cos (π / π) are used as shown in the following equation (6),
8), Cos (3π / 8), Cos (3π / 8) Cos
(Π / 16), Cos (3π / 16), Cos (5π / 1
6), a matrix represented by seven values of Cos (7π / 16).

[0112]

(Equation 13)

On the other hand, the matrix of the reduced inverse discrete cosine coefficient is shown in the following equation (7).

[0114]

[Equation 14]

The matrix of the equation (7) indicates the inverse discrete cosine coefficient of the phase angle π to 98π with the denominator being 16.
In the matrix of equation (7), the coefficient indicated in the top row 0 is the intermediate phase between the coefficients of row 0 and row 1 of the matrix shown in equation (5). In the matrix of equation (7), the coefficient shown in one row is an intermediate phase between the coefficients of the two rows and three rows of the matrix shown in equation (5). In the matrix of this equation (7), the coefficients shown in two rows are intermediate phases between the coefficients of the four rows and five rows of the matrix shown in the above equation (5). Further, the matrix of this equation (7) is 3
The coefficient shown in the row is an intermediate phase between the coefficients of the 6th row and the 7th row of the matrix shown in the above equation (5). When the matrix of Expression (7) is made to correspond to the inverse discrete cosine coefficient having a phase angle of 0 to π / 2, as shown in Expression (8) below,
0, Cos (π / 4), Cos (π / 8), Cos (3π
/ 8) and Cos (π / 8).

[0116]

(Equation 15)

The field mode basis vector reduced inverse discrete cosine transform unit 22 multiplies the two-dimensional determinant as described above in the horizontal and vertical directions, so that one DCT block holding the basis vector A standard resolution image composed of 4 × 4 pixels can be decoded. Further, in the field mode basis vector reduced inverse discrete cosine transform device 22, the phase of each pixel of the image data is such that the vertical phase of each pixel of the top field is 1/2, 5/2,. The vertical phase of each pixel in the field is 1, 3,.... for that reason,
Field mode reduced inverse discrete cosine transform device 1
4 and frame mode reduced inverse discrete cosine transform device 1
5 can decode the image data having the same phase as that of the image data to be decoded, and the image is not deteriorated due to the phase shift.

Next, the two-dimensional matrix processed by the frame mode basis vector reduced inverse discrete cosine transform unit 23 will be described in more detail.

First, the frame mode basis vector reduced inverse discrete cosine transform unit 23 calculates the basis vector D
The two-dimensional matrix by which the CT block is multiplied in the horizontal direction is the same as the two-dimensional matrix used in the field mode basis vector reduced inverse discrete cosine transform device 22.

Subsequently, the basis vector D is converted by the frame mode basis vector reduced inverse discrete cosine transform unit 23.
Only a two-dimensional matrix in which a CT block is multiplied in the vertical direction will be described.

FIGS. 17 to 23 show cosine curves based on first to seventh order basis vectors in the frame mode, and show the phase relationship between ordinary 8-point inverse discrete cosine coefficients and reduced inverse discrete cosine coefficients for basis vectors. Will be described. 17 shows a cosine curve based on a primary basis vector, FIG. 18 shows a cosine curve based on a secondary basis vector, FIG. 19 shows a cosine curve based on a tertiary basis vector, and FIG. 21 shows a cosine curve based on a fifth-order basis vector, FIG. 22 shows a cosine curve based on a sixth-order basis vector, and FIG. 23 shows a cosine curve based on a seventh-order basis vector. The cosine curve is shown. Note that the horizontal axis of each figure shows the phase of the pixel after the inverse discrete cosine transform. In addition, a white circle in the figure indicates 8 in the normal processing in which no reduction is performed.
A point inverse discrete cosine coefficient is shown, and a black circle shows a reduced inverse discrete cosine coefficient for a base vector. It is assumed that white circles are overlapped on all the black circles in FIGS.

As shown by white circles in FIGS. 17 to 23, the eight-point inverse discrete cosine coefficient calculated by the normal processing without performing the reduction is such that the phase on the cosine curve based on the first to seventh order basis vectors is an integer. (0-7). Therefore, the pixel in this DCT block after performing the inverse discrete cosine transform using the 8-point inverse discrete cosine coefficient has a base vector component.

On the other hand, the reduced inverse discrete cosine coefficient for the base vector to be subjected to the reduction processing is the same as the cosine curve on the cosine curve based on the base vector in the normal processing, as shown by the black circles in FIGS. This is the value at the position where the phase of the pixel is 1, 2, 5, and 6, which is the above value. The frame mode basis vector reduced inverse discrete cosine transform device 23 performs a matrix operation in the vertical direction using the reduced inverse discrete cosine coefficient, thereby retaining the same basis vector components as in the normal processing, and performing the reduced inverse discrete cosine transform. Cosine transform can be performed.

Specifically, the matrix of reduced inverse discrete cosine coefficients calculated in the vertical direction by the frame mode basis vector reduced inverse discrete cosine transform unit 23 is shown in the following equation (9).

[0125]

(Equation 16)

The matrix of this equation (9) indicates the inverse discrete cosine coefficient of the phase angle π to 91π with the denominator being 16.
In the matrix of Expression (9), the coefficient indicated in the top row 0 is the same as the coefficient of one row of the matrix expressed by Expression (5). In the matrix of equation (9), the coefficient shown in one row is the same as the coefficient of two rows in the matrix shown in equation (5). Also, in the matrix of this equation (9), the coefficient shown in two rows is the same as the coefficient of five rows of the matrix shown in the above equation (5). In the matrix of this equation (7), the coefficients shown in three rows are the same as the coefficients of six rows in the matrix shown in equation (5).
When the matrix of equation (9) is made to correspond to the inverse discrete cosine coefficient having a phase angle of 0 to π / 2, as shown in the following equation (10), Cos (π / 4) and Cos (3π / 16),
Cos (5π / 16), Cos (3π / 8), Cos (7
π / 16), Cos (π / 16), Cos (π / 8)
It is a matrix represented by this value.

[0127]

[Equation 17]

The frame mode basis vector reduced inverse discrete cosine transform unit 23 performs an operation using the above-described two-dimensional determinant, thereby retaining the basis vector.
One DCT block can decode a standard resolution image composed of 4 × 4 pixels. Further, in the frame mode basis vector reduced inverse discrete cosine transform device 23, the vertical phase of each pixel in the top field is 1/2, 5/2,. The vertical phase of each pixel in the field is 1, 3,.... Therefore, image data having the same phase as that of image data decoded by the field mode reduced inverse discrete cosine transform device 14 and the frame mode reduced inverse discrete cosine transform device 15 can be decoded. Does not occur.

As described above, in the image decoding apparatus 10 according to the first embodiment of the present invention, in the field DCT mode,
4 for each of the top field and bottom field
A standard resolution image is decoded by performing a reduced inverse discrete cosine transform of × 4, and in the frame DCT mode, a standard resolution image is decoded by performing a reduced inverse discrete cosine transform after frame separation. The image decoding apparatus 10 performs the processing in the field DCT mode and the frame DCT mode in this manner, so that the interlacing property of the interlaced scan image is not impaired, and the decoding is performed in the field DCT mode and the frame DCT mode. The phases of the output images can be the same, and the image quality of the output image does not deteriorate.

Furthermore, in the image decoding apparatus 10 according to the first embodiment of the present invention, when the DCT block is a basis vector, the field mode basis vector reduction inverse discrete cosine transform apparatus 22 or the frame mode basis vector reduction is used. The DCT block is decoded using the inverse discrete cosine transform unit 23. As a result, high-frequency coefficients having base vector components can also be decoded, and image quality can be improved. In particular, in the case of MPEG, since the DCT coefficient of an inter macro block is coded so as to be as small as possible to minimize the transmission amount, the DCT coefficient often becomes a base vector. Therefore, in the image decoding apparatus 10 according to the first embodiment, the difference information is not lost due to the base vector, and the image quality is not easily degraded due to block distortion or the like.

(Second Embodiment) Next, an image decoding apparatus according to a second embodiment of the present invention will be described. In the description of the image decoding device according to the second embodiment, the same components as those in the first image decoding device 10 are denoted by the same reference numerals in the drawings, and detailed description thereof will be omitted.

As shown in FIG. 24, an image decoding apparatus 30 according to the second embodiment of the present invention receives a bit stream obtained by compressing a high-resolution image having 1152 effective lines in the vertical direction by MPEG2. Then, this device decodes the input bit stream and reduces it to half the resolution, and outputs a standard resolution image having 576 effective lines in the vertical direction, for example.

The image decoding device 30 is supplied with a bit stream of a compressed high-resolution image, and analyzes the bit stream with a bit stream analyzing device 11.
A variable-length code decoding device 12 that decodes the bit stream that has been subjected to the variable-length coding that assigns a code length according to the frequency of occurrence of data, and an inverse quantization device 13 that applies a quantization step to each coefficient of the DCT block. Have.

Further, the image decoding device 30 sets the field D
A field mode phase-correction reduced inverse discrete cosine transform device 31 for performing a reduced inverse discrete cosine transform on a DCT block subjected to a discrete cosine transform in a CT mode to generate a standard resolution image, and a discrete cosine transform in a frame DCT mode A phase-correction reduced inverse discrete cosine transform device 32 for frame mode that performs reduced inverse discrete cosine transform on the DCT block thus generated to generate a standard resolution image, and a standard resolution image subjected to reduced inverse discrete cosine transform and motion compensation are performed. An addition device 16 for adding the reference image thus obtained, a frame memory 17 for temporarily storing the reference image, and a field mode motion compensation device for performing motion compensation corresponding to the field motion prediction mode on the reference image stored in the frame memory 17 18 and the reference image stored in the frame memory 17 for frame motion prediction. A motion compensation unit for frame mode 19 to the motion compensation corresponding to the mode, the frame memory 1
And an image frame conversion device 33 that outputs image data of a standard resolution for converting the image stored in the image 7 into an image frame and displaying the image on a monitor or the like.

The image decoding apparatus 10 further includes a basis vector discriminating apparatus 21 for discriminating whether or not the coefficients of the DCT block subjected to the quantization step by the inverse quantization apparatus 13 are base vectors, and a field DCT mode. Is a discrete cosine transform and is a base vector DC
A field mode basis vector reduced inverse discrete cosine transform unit 34 for performing a reduced inverse discrete cosine transform on a T block to generate a standard resolution image; and a DC which is a basis vector subjected to discrete cosine transform in a frame DCT mode.
A frame mode basis vector reduced inverse discrete cosine transform unit 35 for performing a reduced inverse discrete cosine transform on the T block to generate a standard resolution image.

The phase-correction inverse discrete cosine transform unit 31 for the field mode is used when the macroblock of the input bit stream is subjected to the discrete cosine transform in the field DCT mode and the DCT block in the macroblock is not a base vector. . The field mode phase-correcting reduced inverse discrete cosine transform device 31 outputs 4 × 8 of all the coefficients of the DCT block in which the 8 × 8 coefficients in the macroblock subjected to the discrete cosine transform in the field DCT mode are indicated. An inverse discrete cosine transform is performed on only the coefficients, in which the phase shift of the pixels in the vertical direction between the top field and the bottom field is corrected. That is,
The inverse discrete cosine transform is performed based on the discrete cosine coefficients of four points in the low frequency range in the horizontal direction, and the phase shift is corrected based on the discrete cosine coefficients of eight points in the vertical direction. More specifically, a phase correction of 1/4 pixel is performed for each pixel in the vertical direction of the top field, and a phase correction of 3/4 pixel is performed for each pixel in the vertical direction of the bottom field. I do. Then, by performing the reduced inverse discrete cosine transform as described above, FIG.
, The vertical phase of each pixel in the top field is 1/4, 9/4,..., And the vertical phase of each pixel in the bottom field is 5/4, 13/4,. A standard resolution image (lower layer) is generated.

The phase-correction reduced inverse discrete cosine transform unit for frame mode 32 is used when the macroblock of the input bit stream is subjected to discrete cosine transform in the frame DCT mode and the DCT block in the macroblock is not a base vector. . The frame mode phase-correcting decimating inverse discrete cosine transform unit 32 outputs the frame D
For a DCT block in which 8 × 8 coefficients are shown in a macroblock subjected to discrete cosine transform in the CT mode, one-block processing or two-block processing, which will be described in detail later, performs vertical processing of a top field and a bottom field. Performs a reduced inverse discrete cosine transform in which the phase shift of the pixel is corrected. Then, an image having the same phase as the phase of the pixel of the standard resolution image generated by the field mode phase-correcting reduced inverse discrete cosine transform device 31 is generated. That is, by performing the reduced inverse discrete cosine transform by one-block processing or two-block processing, the vertical phase of each pixel in the top field as shown in FIG.
.. To generate a standard resolution image (lower layer) in which the vertical phase of each pixel in the bottom field is 5/4, 13/4,.

The field mode basis vector reduced inverse discrete cosine transform unit 34 is used when a macroblock of an input bit stream is subjected to discrete cosine transform in the field DCT mode and a DCT block in the macroblock is a basis vector. Can be Field mode basis vector reduced inverse discrete cosine transform unit 34
Is a DCT showing 8 × 8 coefficients in a macroblock subjected to discrete cosine transform in the field DCT mode
The block is subjected to a two-dimensional matrix multiplication process in the horizontal and vertical directions using the inverse discrete cosine coefficient of the phase on the cosine curve obtained as a result of inverse discrete cosine transform of the coefficients of all frequency components of the DCT block. . This field mode basis vector reduced inverse discrete cosine transform device 3
In No. 4, by performing such a two-dimensional matrix operation, a standard resolution image in which one DCT block is composed of 4 × 4 pixels can be decoded. As shown in FIG. 25, the phase of each pixel of this decoded image data is 1 / vertical phase of each pixel of the top field.
., And the standard resolution image (lower layer) in which the vertical phase of each pixel of the bottom field is 5/4, 13/4,.

The field mode basis vector reduced inverse discrete cosine transform unit 34 processes
Details of the dimensional matrix will be described later.

The frame mode basis vector reduced inverse discrete cosine transform unit 35 is used when a macroblock of an input bit stream is subjected to discrete cosine transform in the frame DCT mode and a DCT block in the macroblock is a basis vector. Can be The frame mode basis vector reduced inverse discrete cosine transform unit 35 applies a discrete cosine in the field DCT mode to the DCT block in which the 8 × 8 coefficients in the macroblock subjected to the discrete cosine transform in the frame DCT mode are indicated. Inverse discrete phase of the cosine curve obtained as a result of performing inverse discrete cosine transform of the coefficients of all the frequency components of the DCT block indicating the 8 × 8 coefficients in the transformed macroblock. Using the cosine coefficient, a two-dimensional matrix multiplication process in the horizontal and vertical directions is performed. The field mode basis vector reduced inverse discrete cosine transform device 22 can decode a standard resolution image in which one DCT block is composed of 4 × 4 pixels by performing such a two-dimensional matrix operation. it can. As shown in FIG. 25, the phase of each pixel of the decoded image data is such that the vertical phase of each pixel of the top field is 1/4, 9/4,. A standard resolution image (lower layer) having a phase in the direction of 5/4, 13/4,... Is generated.

The details of the two-dimensional matrix processed by the frame mode basis vector reduced inverse discrete cosine transform unit 35 will be described later.

The basis vector discriminating device 21 decides whether or not the coefficients of the DCT block which have been subjected to the quantization step by the inverse quantization device 13 are the basis vectors. As a result of the discrimination by the basis vector discriminating device 21, D
If the coefficient of the CT block is not a base vector, the DC quantized by the inverse quantizer 13
The T block is supplied to the field-corrected phase-reduced reduced inverse discrete cosine transform unit 31 for frame mode or the phase-corrected reduced inverse discrete cosine transform unit 32 for frame mode. Also, D
If the coefficient of the CT block is a basis vector, the DC quantized by the inverse quantizer 13
The T block is supplied to a field mode basis vector reduced inverse discrete cosine transform unit 34 or a frame mode basis vector reduced inverse discrete cosine transform unit 35.

The adder 16 includes a field-corrected phase-reduced inverse discrete cosine transform unit 31 for frame mode, a phase-corrected reduced inverse discrete cosine transform unit 32 for frame mode, and a base-vector reduced inverse discrete cosine transform unit 3 for field mode.
If the macroblock subjected to the reduced inverse discrete cosine transform by the 4 or frame mode basis vector reduced inverse discrete cosine transform unit 35 is an intra image, the intra image is stored in the frame memory 17 as it is. Also,
The adder 16 includes a field-corrected phase-corrected reduced inverse discrete cosine transform device 31, a frame-mode phase-corrected reduced inverse discrete cosine transform device 32, a field-mode base vector reduced inverse discrete cosine transform device 34, or a frame-mode base vector reduced. Inverse discrete cosine transform device 3
If the reduced inverse discrete cosine transformed macroblock is an inter image, the inter image
The reference image subjected to motion compensation by the field mode motion compensator 18 or the frame mode motion compensator 19 is synthesized and stored in the frame memory 17.

The field mode motion compensator 18
This is used when the motion prediction mode of the macroblock is the field motion prediction mode. The field mode motion compensator 18 performs an interpolation process on the reference image of the standard resolution image stored in the frame memory 17 with 1/4 pixel accuracy, and performs motion compensation corresponding to the field motion prediction mode. This motion compensation device for field mode 1
The reference image motion-compensated by 8 is supplied to the adding device 16 and is synthesized with the inter image.

The frame mode motion compensator 19 is used when the macroblock motion prediction mode is the frame motion prediction mode. The frame mode motion compensator 19 performs an interpolation process on the reference image of the standard resolution image stored in the frame memory 17 with 1/4 pixel accuracy, and performs motion compensation corresponding to the frame motion prediction mode. The reference image motion-compensated by the frame mode motion compensator 19 is supplied to the adder 16,
It is combined with the inter image.

The image frame conversion device 33 includes the frame memory 17
Is supplied, and the reference image is converted by post-filtering so that the image frame conforms to the standard of the standard resolution television. That is, the image frame conversion device 33 converts the image frame of the high resolution television standard into an image frame of the standard resolution television standard which is reduced to 1/4. Note that this image frame conversion device 33
Since the image stored in the frame memory 17 has no phase shift between the top field and the bottom field, the image frame conversion /
Unlike the phase shift correction device 20, it is not necessary to correct the phase shift of the pixel.

The image decoding apparatus 30 according to the second embodiment of the present invention has the above-described configuration, so that a bit stream obtained by compressing a high-resolution image by MPEG2 can be decoded and the resolution can be reduced to half. And a standard resolution image can be output.

Next, the processing contents of the above-mentioned field mode phase-correcting inverse discrete cosine transform unit 31 will be described.
This will be described in more detail.

As shown in FIG. 26, the phase-correcting decimating inverse discrete cosine transform unit 31 for the field mode includes a bit stream obtained by compression-encoding a high-resolution
It is input in units of CT blocks.

First, in step S21, the discrete cosine coefficient y of this one DCT block (the vertical coefficient among all the discrete cosine coefficients of the DCT block is y
It is shown in the figure as ₁ ~y _8. ) Is subjected to an 8 × 8 inverse discrete cosine transform (IDCT 8 × 8). By performing an inverse discrete cosine transform, 8 × 8 decoded pixel data x (vertical pixel data among all the pixel data of the DCT block is shown as x ₁ to x ₈ in the drawing) is obtained. Can be.

Subsequently, in step S22, this 8
The pixel data of × 8, performs conversion closed within DCT blocks by the phase correcting filter matrix 4 × 8, 'x vertical pixel data among (all pixel data' pixel data x that phase correction _1, x ' ₂ , x' ₃ , x ' ₄ are shown in the figure).

By performing the above-described steps S21 to S22, the field mode phase-correcting decimating inverse discrete cosine transform apparatus 31 generates an image having no pixel phase shift between the top field and the bottom field. can do.

In the field mode phase-correcting reduced inverse discrete cosine transform unit 31, as shown in FIG.
The above processing may be calculated using one matrix (4 × 8 phase correction IDCT matrix).

Next, FIG. 28 shows a procedure for designing a 4 × 8 phase-corrected IDCT matrix operated by the field mode phase-corrected inverse discrete cosine transform unit 31. The 4 × 8 phase-corrected IDCT matrix will be described. I do. This 4 × 8 phase correction IDCT matrix is created by polyphase decomposition of a prototype filter.

Here, in the image decoding apparatus 30, FIG.
A high-resolution image having frequency characteristics as shown in FIG.
The signal band shown in FIG. 9 (b) is down-decoded into a standard resolution image having a resolution of 1/2 of the frequency characteristic halved by the low-pass filter. For this reason, the frequency characteristics required for the prototype filter are such that the pixel value of 1/4 phase of the standard resolution image can be obtained as shown in FIG.
As shown in FIG. 9C, the frequency characteristics are obtained by performing the oversampling by four times.

First, in step S31, Nyquist frequency and lower frequencies are divided into {(N-1) / 2} at equal intervals, and a gain list is created from the frequency samples. For example, as shown in FIG. 30, a frequency lower than the Nyquist frequency is divided into (57-1) / 2 = 28 at equal intervals to create 29 gain lists.

Subsequently, in step S32, 57 impulse responses are created by the frequency sampling method. That is, the 29 discrete gain lists are subjected to inverse discrete Fourier transform to create impulse responses of 57 FIR filters. FIG. 31 shows these 57 impulse responses.

Subsequently, in step S33, the impulse response is multiplied by a window function to generate filter coefficients c1 to c57 of 57 taps.

The filter created in step S33 is a prototype filter.

Subsequently, in step S34, the prototype filter having 57 filter coefficients c1 to c57 is polyphase-decomposed, and only 14 filter coefficients c'1 to c'14 having 1/4 phase correction characteristics are obtained. And create a polyphase filter.

Here, the polyphase filter corresponds to FIG.
As shown in FIG. 2, the input signal was oversampled by N times, and polyphase decomposition was performed to extract pixels at N pixel intervals from the signal obtained by oversampling, and the input signal was shifted by 1 / N phase. This is a filter that outputs a signal. For example, in order to obtain a signal whose phase is shifted by 1/4 with respect to the input signal, as shown in FIG. 33, the input signal is oversampled by a factor of 4 and a signal shifted by 1/4 phase from this signal can be extracted. I just need.

Specifically, the 14 filter coefficients c'1 to c'14 created from the prototype filters c1 to c57 having 57 coefficients are obtained by, for example, the following equation (1
The coefficients are as shown in 1).

[0163]

(Equation 18)

After creating the polyphase filter in this manner, the design process is divided into a 4 × 8 phase correction IDCT matrix for the top field and a 4 × 8 phase correction IDCT matrix for the bottom field.

First, when creating a 4 × 8 phase correction IDCT matrix for the top field, in step S35, the 14 filters subjected to polyphase decomposition are set so that the filter coefficients have a ４ phase correction characteristic. Coefficient c '
From 1 to c′14, the group delay is ４, 9/4, 17 /
Take out eight coefficients which are 4, 25/4 phase, 4 × 8
Create a phase correction filter matrix. FIG. 34 shows the 4 × 8 phase correction filter thus created.

For example, from the 14 filter coefficients c′1 to c′14 of the above equation (11), coefficients as shown in the following equation (12) are extracted.

[0167]

[Equation 19]

When a 4 × 8 phase correction filter matrix is obtained from the coefficients of the equation (12), the matrix becomes as shown in the following equation (13).

[0169]

(Equation 20)

When the 4 × 8 phase correction filter matrix shown in the equation (13) is normalized, a matrix shown in the following equation (14) is obtained.

[0171]

(Equation 21)

Then, in step S36, 8 × 8
Is multiplied by this 4 × 8 phase correction filter matrix to obtain a 4 × 8 phase correction I
Create a DCT matrix.

4 × 8 obtained by multiplying the 8 × 8 IDCT matrix by the 4 × 8 phase correction filter shown in the above equation (14)
The phase correction IDCT matrix is a matrix as shown in the following equation (15).

[0174]

(Equation 22)

On the other hand, when preparing a 4 × 8 phase correction IDCT matrix for the bottom field, in step S37, the 14 filters subjected to polyphase decomposition are set so that the filter coefficients have a 3/4 phase correction characteristic. Coefficient c '
1 to c ′ 14 are reversed left and right.

Subsequently, in step S38, from the 14 filter coefficients c'1 to c'14 which have been inverted left and right,
Eight coefficients having group delays of 3/4, 11/4, 19/4, and 27/4 are extracted, and a 4 × 8 phase correction filter matrix is created.

Then, in step S39, 8 × 8
Is multiplied by the 4 × 8 phase correction filter matrix to obtain a 4 × 8 phase correction I for the bottom field.
Create a DCT matrix.

As described above, steps S31 to S3
9, the field mode phase-correcting reduced inverse discrete cosine transform unit 31 performs the operation.
A × 8 phase correction IDCT matrix can be created.

As described above, in the field mode phase-correcting reduced inverse discrete cosine transform unit 31, the 4 × 8 phase-corrected IDCT matrix and the coefficients of the DCT block subjected to the discrete cosine transform in the input field DCT mode are used. Can be decoded into a standard resolution image having no phase shift between the top field and the bottom field. That is, in the field-corrected phase-reduced inverse discrete cosine transform device 31 for the field mode, the vertical phase of each pixel in the top field is 1/4, 9/4,... As shown in FIG. The vertical phase of each pixel is 5/4, 13 /
4... Can be generated.

Next, the processing contents of the frame mode phase-correcting decimating inverse discrete cosine transform unit 32 will be described in more detail.

The frame mode phase-correcting decimating inverse discrete cosine transform unit 32 can perform one or both of the following one-block processing and two-block processing. If necessary, one block processing or two block processing may be switched and used, or only one of the processing may be performed.

First, the one-block processing will be described.
FIG. 35 shows a diagram for explaining the contents of one block processing.

As shown in FIG. 35, a bit stream obtained by compression-encoding a high-resolution image is converted into one DC
It is input in units of T blocks.

First, in step S41, the discrete cosine coefficient y of this one DCT block is set to 8 ×
8 is performed. Subsequently, step S4
In step 2, the 8 × 8 pixel data is field-separated. Subsequently, in step S43, a 4 × 4 discrete cosine transform is performed on each of the two pixel blocks separated by the field. Subsequently, in step S44, the high frequency components of the discrete cosine coefficient z of each pixel block are thinned out to obtain a pixel block composed of 2 × 2 discrete cosine coefficients. The above steps S41 to S
The processing up to 44 is the same as the processing from step S1 to step S4 in the one-block processing shown in FIG.

Subsequently, in step S45, 1/4 of the pixel block corresponding to the top field
Inverse discrete cosine transform is performed by using a 2 × 4 phase correction IDCT matrix for correcting the phase of the pixels and correcting the phase shift of the pixels in the vertical direction. For the pixel block corresponding to the bottom field, inverse discrete cosine transform in which the phase shift of pixels in the vertical direction has been corrected using a 2 × 4 phase correction IDCT matrix for phase correction of ／ pixels is performed. Do. By performing the reduced inverse discrete cosine transform as described above,
2 × 2 pixel data x ′ (vertical pixel data of all pixel data of the pixel block corresponding to the top field is shown in the figure as x ′ ₁ , x ′ ₃ , and the pixel corresponding to the bottom field shown in FIG vertical pixel data of all the pixel data of block x as _'2, x' _4.) can be obtained. This pixel data x '
Is that the vertical phase of each pixel in the top field is 1 /
4, 9/4, and a standard resolution image (lower layer) in which the vertical phase of each pixel in the bottom field is 5/4, 13/4. Note that this 2 × 4 phase correction I
The details of the method of designing the DCT matrix will be described later.

Subsequently, in step S46, the pixel data of the pixel block corresponding to the top field and the pixel data of the image block of the bottom field are frame-synthesized. The processing in step S46 is performed by the processing shown in FIG.
This is the same as the processing in step S6 in the block processing.

By performing the processing in steps S41 to S46, the frame mode phase-correcting reduced inverse discrete cosine transform apparatus 32 can generate an image having no phase shift between pixels. Further, it is possible to generate an image having no phase shift from the image decoded by the field mode phase-correcting reduced inverse discrete cosine transform device 31.

In the frame mode phase-correcting decimating inverse discrete cosine transform unit 32, the processing from step S41 to step S46 may be performed using one matrix.

Next, the design procedure of the 2 × 4 phase-corrected IDCT matrix to be operated in step S45 of the frame mode phase-corrected reduced inverse discrete cosine transform unit 32 is shown in FIG.
The 2 × 8 phase correction IDCT matrix will be described.

First, in step S51, the frequency below the Nyquist frequency is divided into {(N-1) / 2} at equal intervals, and a gain list is created from the frequency samples. For example, as shown in FIG. 37, the frequencies below the Nyquist frequency are divided into (25-1) / 2 = 12 at equal intervals to create 13 gain lists.

Subsequently, in step S52, 25 impulse responses are created by the frequency sampling method. That is, the 13 gain lists are subjected to inverse discrete Fourier transform to create impulse responses of 25 FIR filters. FIG. 38 shows these 25 impulse responses.

Subsequently, in step S53, the impulse response is multiplied by a window function to generate 25 tap filter coefficients c1 to c25.

The filter created in step S53 is a prototype filter.

Subsequently, in step S54, the prototype filter having 25 filter coefficients c1 to c25 is polyphase-decomposed, and only six filter coefficients c'1 to c'6 having quarter-phase correction characteristics are obtained. And create a polyphase filter.

More specifically, the 14 filter coefficients c ′ 1 to c ′ 6 created from the prototype filters c 1 to c 25 having 57 coefficients are obtained by, for example, the following equation (16).
The coefficient becomes as shown by.

[0196]

(Equation 23)

After the polyphase filter is created, the design process is divided into a 2 × 4 phase correction IDCT matrix for the top field and a 2 × 4 phase correction IDCT matrix for the bottom field.

First, when preparing a 2 × 4 phase correction IDCT matrix for the top field, in step S 55, the group delay is set to 1 based on the six filter coefficients c ′ 1 to c ′ 6 subjected to polyphase decomposition. Two coefficients are respectively extracted so as to have / 4 and 9/4 phases, and a 2 × 4 phase correction filter matrix is created. 2 created in this way
FIG. 39 shows a × 4 phase correction filter.

For example, from the six filter coefficients c'1 to c'6 in the above equation (16), coefficients as shown in the following equation (17) are extracted.

[0200]

(Equation 24)

When a 2 × 4 phase correction filter matrix is obtained from the coefficients of the equation (17), the matrix becomes as shown in the following equation (18).

[0202]

(Equation 25)

When the 2 × 4 phase correction filter matrix shown in the equation (18) is normalized, the matrix becomes as shown in the following equation (19).

[0204]

(Equation 26)

Then, in step S56, 4 × 4
Is multiplied by the 2 × 4 phase correction filter matrix to obtain a 2 × 4 phase correction I for the top field.
Create a DCT matrix.

A 2 × 4 IDCT matrix multiplied by a 2 × 4 phase correction filter shown in the above equation (19)
The phase correction IDCT matrix is a matrix as shown in the following equation (20).

[0207]

[Equation 27]

On the other hand, when preparing a 2 × 4 phase correction IDCT matrix for the bottom field, in step S57, the six filters subjected to polyphase decomposition are set so that the filter coefficients have a 3/4 phase correction characteristic. Coefficient c'1
To c′6 are reversed left and right.

Subsequently, in step S58, two coefficients are respectively determined from the six filter coefficients c'1 to c'6 which have been inverted left and right so that the group delay becomes 3/4 and 11/4 phase. Then, a 2 × 4 phase correction filter matrix is created.

In step S59, 4 × 4
Is multiplied by the 2 × 4 phase correction filter matrix to obtain a 2 × 4 phase correction I for the bottom field.
Create a DCT matrix.

As described above, steps S51 to S
By performing each of the processes 59, it is possible to create a 2 × 4 phase-corrected IDCT matrix for which the frame-mode phase-corrected decimating inverse discrete cosine transform device 32 performs the operation in step S45.

Next, the two-block processing will be described. FIG. 40 shows a diagram for explaining the contents of the two-block processing.

As shown in FIG. 40, a bit stream obtained by compression-encoding a high-resolution image includes two DC
It is input in units of T blocks. For example, when a macroblock is composed of a DCT block of four luminance components and a DCT block of two color difference components, two vertically adjacent DCT blocks are input. For example,
When the macro block is configured as shown in FIG. 5, the DCT blocks 0 and D of the luminance component (Y) are used.
CT block 2 is input as a pair, and DCT
Block 1 and DCT block 3 are input as a pair.

First, in step S61, two D
An 8 × 8 inverse discrete cosine transform is independently performed on the discrete cosine coefficient y of the CT block. By performing the inverse discrete cosine transform, 8 × 8 decoded pixel data x can be obtained. Subsequently, in step S62, two 8 × 8 pixel data are field-separated. Subsequently, in step S63, each of the two 8 × 8 pixel blocks separated by the field is divided into 8 × 8 pixel blocks.
8 discrete cosine transform. Subsequently, step S64
, The high-frequency component of the discrete cosine coefficient z of the pixel block corresponding to the top field obtained by performing the 8 × 8 discrete cosine transform is thinned out to obtain a pixel block including 4 × 4 discrete cosine coefficients. . Further, the high-frequency component of the discrete cosine coefficient z of the pixel block corresponding to the bottom field obtained by performing the 8 × 8 discrete cosine transform is thinned out to be a pixel block composed of 4 × 4 discrete cosine coefficients.

The above steps S61 to S64
The processing up to is the same as the processing from step S11 to step S14 in the two-block processing shown in FIG.

Subsequently, in step S65, for the pixel block of the top field, a 4 × 8 phase correction IDCT matrix for performing phase correction of 画素 pixel is used.
An inverse discrete cosine transform in which the phase shift of the pixel in the vertical direction is corrected is performed. For the pixel block of the bottom field, inverse discrete cosine transform in which the phase shift of pixels in the vertical direction has been corrected is performed using a 4 × 8 phase correction IDCT matrix for phase correction of 、 ３ pixels. By performing the above-described reduced inverse discrete cosine transform, 4 × 4 pixel data x ′ (vertical pixel data of all pixel data of the pixel block corresponding to the top field is x ′ ₁ ,
In the figure, x ′ ₃ , x ′ ₅ , x ′ ₇ are shown in the figure, and vertical pixel data among all the pixel data of the pixel block corresponding to the bottom field are x ′ ₂ , x ′ ₄ , x ′ ₆ ,
As x _'8 shown in FIG. ) Can be obtained. In the pixel data x ', the vertical phase of each pixel in the top field is 1/4, 9/4,..., And the vertical phase of each pixel in the bottom field is 5/4, 13/4,.
Generate a standard resolution image (lower layer).
Note that the design method of this 4 × 8 phase correction IDCT matrix is as follows.
The matrix is the same as the matrix calculated by the above-described field mode phase-correcting reduced inverse discrete cosine transform unit 31.

Subsequently, in step S66, the pixel data of the pixel block corresponding to the top field and the pixel data of the pixel block corresponding to the bottom field are alternately frame-combined line by line in the vertical direction.
A DCT block that has been subjected to a reduced inverse discrete cosine transform composed of 8 × 8 pixel data is generated.

By performing the two block processes of steps S61 to S66 described above, the frame mode phase-correcting reduced inverse discrete cosine transform device 32 can generate an image having no phase shift between pixels. Further, it is possible to generate an image having no phase shift from the image decoded by the field mode phase-correcting reduced inverse discrete cosine transform device 31.

In the frame mode phase-correcting decimating inverse discrete cosine transform unit 32, the processing from step S61 to step S66 may be performed using one matrix.

Next, the two-dimensional matrix in which the DCT block of the base vector is multiplied in the horizontal and vertical directions by the field mode base vector reduced inverse discrete cosine transform unit 34 will be described in further detail.

First, the two-dimensional matrix by which the DCT block of the basis vector is multiplied in the horizontal direction by the field mode basis vector reduced inverse discrete cosine transform unit 34 is the field mode basis vector according to the first embodiment. This is the same as the two-dimensional matrix used in the basis vector reduced inverse discrete cosine transform unit 22.

Hereinafter, only a two-dimensional matrix in which the DCT block of the basis vector is multiplied in the vertical direction by the field mode basis vector reduced inverse discrete cosine transform unit 34 will be described.

FIGS. 41 to 47 show cosine curves based on first to seventh order basis vectors in the field mode, and show the phase relationship between ordinary 8-point inverse discrete cosine coefficients and reduced inverse discrete cosine coefficients for basis vectors. Will be described.
41 shows a cosine curve based on a primary basis vector, FIG. 42 shows a cosine curve based on a secondary basis vector, FIG. 43 shows a cosine curve based on a tertiary basis vector, and FIG. 45 shows a cosine curve based on a fifth-order basis vector, FIG. 46 shows a cosine curve based on a sixth-order basis vector, and FIG. 47 shows a cosine curve based on a seventh-order basis vector. The cosine curve is shown. Note that the horizontal axis of each figure shows the phase of the pixel after the inverse discrete cosine transform. In addition, a white circle in the figure indicates 8 in the normal processing in which no reduction is performed.
Indicates the inverse discrete cosine coefficient of the point, and the black square indicates the reduced inverse discrete cosine coefficient for the basis vector in the top field is 1
係数 indicates a coefficient shifted by ４, and a black triangle indicates a coefficient shifted by ／ from the reduced inverse discrete cosine coefficient for the base vector in the bottom field.

As shown by white circles in FIGS. 41 to 47, the eight-point inverse discrete cosine coefficient calculated by the normal processing without performing the reduction is such that the phase on the cosine curve based on the first to seventh order basis vectors is an integer. (0-7). Therefore, the pixel in this DCT block after performing the inverse discrete cosine transform using the 8-point inverse discrete cosine coefficient has a base vector component.

On the other hand, the reduced inverse discrete cosine coefficient for the base vector of the top field to be reduced is 1 /
The coefficients shifted by four phases are values on the same cosine curve as those on the cosine curve based on the basis vectors of the normal processing, as indicated by black squares in FIGS. 41 to 47, and the phase of the pixel is 0.25. , 2.25, 4.25, and 6.25. The field mode basis vector reduced inverse discrete cosine transform device 34 performs a matrix operation in the vertical direction using the reduced inverse discrete cosine coefficient,
The reduced inverse discrete cosine transform can be performed while holding the same basis vector components as in the normal processing.

Further, the coefficient obtained by shifting the reduced inverse discrete cosine coefficient for base vector of the bottom field to be reduced by 3/4 phase, as shown by black triangles in FIGS.
The value on the cosine curve is the same as the value on the cosine curve based on the basis vector of the normal processing, and the phase of the pixel is 0.
75, 2.75, 4.75, and 6.75. The field mode basis vector reduced inverse discrete cosine transform unit 34 performs a matrix operation in the vertical direction using the reduced inverse discrete cosine coefficient, thereby retaining the same basis vector components as in the normal processing, and performing the reduced inverse discrete cosine transform. Cosine transform can be performed.

The matrix of the reduced inverse discrete cosine coefficient calculated by the field mode basis vector reduced inverse discrete cosine transform unit 34 will be specifically described below.

First, a matrix of reduced inverse discrete cosine coefficients in the top field having been subjected to 1/4 phase correction is shown in the following equation (21).

[0229]

[Equation 28]

The matrix of this equation (21) indicates the inverse discrete cosine coefficient having a phase angle of 3π to 189π with the denominator being 32. In the matrix of this equation (21), the coefficient shown in the top row 0 is the quarter phase of the phase angle of the coefficient of the row 0 and the row 1 of the matrix shown in the above equation (5). ing. Also,
In the matrix of this equation (21), the coefficient shown in one row is １ ／ phase of the phase angle of the coefficients of the two rows and three rows of the matrix shown in the above equation (5). In the matrix of this equation (21), the coefficient shown in two rows is １ ／ phase of the phase angle of the coefficients of the four rows and five rows of the matrix shown in equation (5). . Further, in the matrix of this equation (21), the coefficient shown in three rows is １ ／ phase of the phase angle of the coefficient of six rows and seven rows of the matrix shown in the above equation (5). . Then, when the matrix of this equation (21) is made to correspond to the inverse discrete cosine coefficient having a phase angle of 0 to π / 2, the matrix becomes as shown in the following equation (22).

[0231]

(Equation 29)

The following equation (23) shows the matrix of the reduced inverse discrete cosine coefficient in the bottom field after the 3/4 phase correction.

[0233]

[Equation 30]

The matrix of this equation (23) indicates the inverse discrete cosine coefficient having a phase angle of 5π to 203π with the denominator being 32. In the matrix of this equation (23), the coefficient shown in the top row 0 is the ３ phase of the phase angle of the coefficient of the row 0 and the row 1 of the matrix shown in the above equation (5). ing. Also,
In the matrix of the equation (21), the coefficient shown in one row is the 位相 phase of the phase angle of the coefficients of the two rows and the three rows of the matrix shown in the above equation (5). In the matrix of this equation (21), the coefficient shown in two rows is 3/4 of the phase angle of the coefficients of the four rows and five rows of the matrix shown in equation (5). . Also, in the matrix of this equation (21), the coefficient shown in three rows is 3/4 of the phase angle of the coefficients of the six rows and seven rows of the matrix shown in equation (5). . Then, when the matrix of this equation (21) is made to correspond to the inverse discrete cosine coefficient having a phase angle of 0 to π / 2, the matrix becomes as shown in the following equation (24).

[0235]

(Equation 31)

The field mode basis vector reduced inverse discrete cosine transform unit 34 multiplies the two-dimensional determinant as described above in the vertical direction, so that one DCT block holding the basis vector is 4 × 4. Can be decoded. further,
The field-mode basis vector reduced inverse discrete cosine transform unit 34 determines that the vertical phase of each pixel in the top field is 1/4, 9 /
.., And a standard resolution image in which the vertical phase of each pixel in the bottom field is 5/4, 13/4,. Therefore, it is possible to decode the image data having the same phase as the phase of the image data decoded by the field mode phase-correction reduced inverse discrete cosine transform device 31 and the frame mode phase-corrected reduced inverse discrete cosine transform device 32, The image does not deteriorate due to the phase shift.

Next, the two-dimensional matrix processed by the frame mode basis vector reduced inverse discrete cosine transform unit 35 will be described in more detail.

The frame mode basis vector reduced inverse discrete cosine transform unit 35 converts the basis vector DC
The two-dimensional matrix by which the T block is multiplied in the horizontal and vertical directions is calculated by using the field mode basis vector reduced inverse discrete cosine transform unit 22 according to the first embodiment.
Is the same as the two-dimensional matrix used in. For this reason, the values on the cosine curve are the same as those on the cosine curve based on the base vector of the normal processing, and the phases of the pixels in the vertical direction are 0.5, 2.5, 4.5, and 6.5. This is the position value. The frame mode basis vector reduced inverse discrete cosine transform device 35 performs a matrix operation in the horizontal and vertical directions using the reduced inverse discrete cosine coefficient, thereby holding the basis vector components similar to those in the normal processing and performing the reduction. Inverse discrete cosine transform can be performed.

The frame mode basis vector reduced inverse discrete cosine transform unit 35 performs an operation using the two-dimensional determinant as described above, so that
One DCT block can decode a standard resolution image composed of 4 × 4 pixels. Further, the frame mode basis vector reduced inverse discrete cosine transform device 35 converts the image data decoded by the field mode phase corrected reduced inverse discrete cosine transform device 31 and the frame mode phase corrected reduced inverse discrete cosine transform device 32. Image data having the same phase as the phase can be decoded, and the image does not deteriorate due to the phase shift.

As described above, in the image decoding device 30 according to the second embodiment of the present invention, in the field DCT mode,
4 for each of the top field and bottom field
A standard resolution image is decoded by performing a reduced inverse discrete cosine transform of × 4 and performing a phase correction, and in the frame DCT mode, a standard resolution image is obtained by performing a frame separation and performing a reduced inverse discrete cosine transform and performing a phase correction. Decrypt.
In this image decoding device 30, the field DC
Since the processing is performed in each of the T mode and the frame DCT mode, the phase shift between the top field and the bottom field caused when the reduced inverse discrete cosine transform is performed without impairing the interlace property of the interlaced image. And the image quality of the output image is not degraded. That is, in the image decoding device 30, when outputting the decoded image stored in the frame memory 17, there is no need to perform the phase correction, thereby simplifying the process and preventing the image quality from deteriorating.

In the image decoding apparatus 30 according to the second embodiment of the present invention, when the DCT block is a basis vector, the field mode basis vector reduction inverse discrete cosine transform apparatus 34 or the frame mode basis vector reduction is used. The DCT block is decoded by using the inverse discrete cosine transform unit 35. As a result, high-frequency coefficients having base vector components can also be decoded, and image quality can be improved. In particular, in the case of MPEG, since the DCT coefficient of an inter macro block is coded so as to be as small as possible to minimize the transmission amount, the DCT coefficient often becomes a base vector. Therefore, in the image decoding device 30 according to the second embodiment, the difference information is not lost due to the base vector, and the image quality is not easily degraded due to block distortion or the like.

Although the image decoding apparatuses according to the first and second embodiments of the present invention have been described above, the data processed by the present invention is not limited to the MPEG2 system image data. In other words, what kind of compressed image data of the first resolution is subjected to predictive coding by performing motion prediction in predetermined pixel block units and compression coding by performing orthogonal transformation in predetermined pixel block units. It may be data. For example, compressed image data using a wavelet method or the like may be used.

[0243]

According to the present invention, when the coefficients of the orthogonal transform block are base vectors, the same curve is obtained as the curve obtained by inverse orthogonal transform of the coefficients of all the frequency components of the orthogonal transform block. Using the inverse orthogonal transform coefficient of the low frequency component of the obtained phase, the inverse orthogonal transform is performed on the coefficient of the low frequency component among the coefficients of the orthogonal transform block. Then, in the present invention, the moving image data having the second resolution lower than the first resolution and holding the base vector component is output.

As a result, according to the present invention, the amount of calculation and storage capacity required for decoding can be reduced, and the moving image data of the second resolution to be output can be output without impairing the interlacing property of the interlaced image. Can be eliminated. In the present invention,
The image quality of the moving image data of the second resolution can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of an image decoding device according to a first embodiment of the present invention.

FIG. 2 is a diagram for explaining a vertical pixel phase of a reference image stored in a frame memory of the image decoding device according to the first embodiment.

FIG. 3 is a diagram for explaining the contents of one block process of the reduced inverse discrete cosine transform device for frame mode of the image decoding device according to the first embodiment.

FIG. 4 is a diagram for explaining the contents of two-block processing of the reduced inverse discrete cosine transform unit for frame mode of the image decoding device according to the first embodiment.

FIG. 5 is a diagram illustrating a DCT block of a luminance component and a chrominance component in a macroblock of 420 format.

FIG. 6 is a diagram showing an operation flow when Wang's algorithm is applied to the processing of the field mode reduced inverse discrete cosine transform device of the image decoding device of the first embodiment.

FIG. 7 is a diagram illustrating a calculation flow when Wang's algorithm is applied to one block processing of the frame mode reduced inverse discrete cosine transform device of the image decoding device according to the first embodiment.

FIG. 8 is a diagram showing D divided into regions in order to determine a basis vector.
It is a figure showing a CT block.

FIG. 9 is a flowchart for determining whether a coefficient in a DCT block is a base vector.

FIG. 10 is based on a first-order basis vector obtained by plotting reduced inverse discrete cosine coefficients for a basis vector calculated by the field mode basis vector reduced inverse discrete cosine transform device of the image decoding device according to the first embodiment. It is a figure showing a cosine curve.

FIG. 11 is a diagram showing a cosine curve based on a quadratic basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 12 is a diagram showing a cosine curve based on a cubic basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 13 is a diagram showing a cosine curve based on a fourth-order basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 14 is a diagram showing a cosine curve based on a fifth-order basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 15 is a diagram showing a cosine curve based on a sixth-order basis vector, in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 16 is a diagram showing a cosine curve based on a seventh-order basis vector, in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 17 is based on a first-order basis vector in which reduced inverse discrete cosine coefficients for basis vectors calculated by the frame mode basis vector reduced inverse discrete cosine transform device of the image decoding apparatus according to the first embodiment are plotted. It is a figure showing a cosine curve.

FIG. 18 is a diagram showing a cosine curve based on a quadratic basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 19 is a diagram showing a cosine curve based on a cubic basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 20 is a diagram showing a cosine curve based on a fourth-order basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 21 is a diagram showing a cosine curve based on a fifth-order basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 22 is a diagram showing a cosine curve based on a sixth-order basis vector, in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 23 is a diagram showing a cosine curve based on a seventh-order basis vector, in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 24 is a block diagram illustrating an image decoding device according to a second embodiment of this invention.

FIG. 25 is a diagram for explaining a phase of a pixel in a vertical direction of a reference image stored in a frame memory of the image decoding device according to the second embodiment.

FIG. 26 is a diagram for describing the processing contents of a field-correction phase-reduced inverse discrete cosine transform apparatus for a field mode of the image decoding apparatus according to the second embodiment.

FIG. 27 is a diagram for explaining the processing contents of the field-correction phase-reduced inverse discrete cosine transform apparatus for the case of performing processing using one matrix.

FIG. 28 is a flowchart for explaining a design procedure of a 4 × 8 phase-corrected IDCT matrix operated by the above-described field-mode phase-corrected reduced inverse discrete cosine transform apparatus.

FIG. 29 is a diagram for explaining frequency characteristics of a prototype filter required for designing the 4 × 8 phase correction IDCT matrix.

FIG. 30: 以下 (N−
FIG. 2 is a diagram for explaining a gain list created from 1) / 2｝ divided frequency samples.

FIG. 31 is a diagram for explaining an impulse response created by performing an inverse discrete Fourier transform on the gain list.

FIG. 32 is a diagram for describing a polyphase filter.

FIG. 33 is a diagram for explaining a polyphase filter that outputs a signal shifted by 1/4 phase from an input signal.

FIG. 34 is a diagram for explaining a 4 × 8 phase-corrected IDCT matrix operated by the above-described field-mode phase-corrected reduced inverse discrete cosine transform apparatus.

FIG. 35 illustrates one example of the frame mode phase-corrected reduced inverse discrete cosine transform apparatus of the image decoding apparatus according to the second embodiment.
It is a figure for explaining the contents of block processing.

FIG. 36 shows a 2 × 4 phase-corrected IDC operated by a frame-corrected phase-reduced inverse discrete cosine transform apparatus for frame mode.
9 is a flowchart for explaining a T matrix design procedure.

FIG. 37: At equal intervals below the Nyquist frequency, Δ (N−
FIG. 2 is a diagram for explaining a gain list created from 1) / 2｝ divided frequency samples.

FIG. 38 is a diagram illustrating an impulse response created by performing an inverse discrete Fourier transform on the gain list.

FIG. 39 shows a 2 × 4 phase correction I calculated by the frame mode phase correction reduced inverse discrete cosine transform device.
FIG. 3 is a diagram for explaining a DCT matrix.

FIG. 40 illustrates a second example of the frame mode phase-correction reduced inverse discrete cosine transform device of the image decoding device according to the second embodiment.
It is a figure for explaining the contents of block processing.

FIG. 41 is based on a first-order basis vector in which reduced inverse discrete cosine coefficients for basis vectors calculated by the field mode basis vector reduced inverse discrete cosine transform device of the image decoding apparatus according to the first embodiment are plotted. It is a figure showing a cosine curve.

FIG. 42 is a diagram showing a cosine curve based on a quadratic basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 43 is a diagram showing a cosine curve based on a cubic basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 44 is a diagram showing a cosine curve based on a fourth-order basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 45 is a diagram showing a cosine curve based on a fifth-order basis vector in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 46 is a diagram showing a cosine curve based on a sixth-order basis vector, in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 47 is a diagram showing a cosine curve based on a seventh-order basis vector, in which the reduced inverse discrete cosine coefficients for the basis vector are plotted.

FIG. 48 is a block diagram showing a conventional first down decoder.

FIG. 49 is a block diagram showing a second conventional down decoder.

FIG. 50 is a block diagram showing a third conventional down decoder.

FIG. 51 is a block diagram of a conventional image decoding device.

FIG. 52 shows a field DCT of the conventional image decoding apparatus.
FIG. 10 is a diagram for describing a reduced inverse discrete cosine transform process in a mode.

FIG. 53 shows a field DCT of the conventional image decoding apparatus.
FIG. 10 is a diagram for describing a reduced inverse discrete cosine transform process in a mode.

FIG. 54 is a diagram for describing linear interpolation processing in the field motion prediction mode of the conventional image decoding device.

FIG. 55 is a diagram for describing linear interpolation processing in the frame motion prediction mode of the conventional image decoding device.

FIG. 56 shows a field DCT of the conventional image decoding apparatus.
FIG. 4 is a diagram for explaining a phase of a pixel obtained as a result of a mode.

FIG. 57 is a diagram for describing a phase of a pixel obtained as a result of the frame DCT mode of the conventional image decoding device.

[Explanation of symbols]

10, 30 image decoding device, 14 reduced inverse discrete cosine transform device for field mode, 15 reduced inverse discrete cosine transform device for frame mode, 16 adder, 17
Frame memory, 18 field mode motion compensator, 19 frame mode motion compensator, 20 picture frame conversion / phase shift correction device, 21 basis vector discriminator, 22, 34 field mode basis vector reduced inverse discrete cosine transform device, 23, 35 Frame mode basis vector reduced inverse discrete cosine transform device, 31 Field mode phase corrected reduced inverse discrete cosine transform device, 32 Frame mode phase corrected reduced inverse discrete cosine transform device, 33 Frame conversion device

Continued on the front page (72) Inventor Kazushi Sato 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Satoshi Mitsuhashi 6-35, 7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Incorporated (72) Inventor Shozo Goseki 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Corporation (72) Inventor Naofumi Yanagihara 6-35, Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Stock Company F term (reference) 5C059 KK01 LA05 LB05 LB18 LC08 LC09 MA03 MA05 MA23 MC22 PP04 SS05 TA05 TA49 TB08 TB10 TC01 TC04 TC27 TD11 TD13 UA05 UA33

Claims (6)

[Claims] 1. A predetermined pixel block (macro block)
From the compressed image data of the first resolution, which has been subjected to prediction coding by performing motion prediction in units and compression coding by performing orthogonal transformation in units of predetermined pixel blocks (orthogonal transformation blocks), An image decoding apparatus for decoding moving image data having a second resolution lower than the resolution, wherein an inverse orthogonal transform is performed on a coefficient of a low frequency component among coefficients of an orthogonal transform block of the compressed image data that has been subjected to the orthogonal transform. An inverse orthogonal transforming unit; an adding unit for adding the compressed image data subjected to the inverse orthogonal transform by the inverse orthogonal transforming unit and the motion-compensated reference image data to output moving image data having a second resolution. A storage unit for storing moving image data output from the adding unit as reference image data; and a macroblock of the reference image data stored in the storage unit. A motion compensating means for performing motion compensation by using the inverse orthogonal transform means. When the coefficients of the orthogonal transform block are base vectors, the inverse orthogonal transform means obtains a result of inverse orthogonal transform of coefficients of all frequency components of the orthogonal transform block. An image decoding apparatus that performs an inverse orthogonal transform on a given curve using an inverse orthogonal transform coefficient of a low-frequency component having a phase on the same curve. 2. The orthogonal transform unit according to claim 1, wherein the orthogonal transform block of the compressed image data subjected to the orthogonal transform by the orthogonal transform method (field orthogonal transform mode) corresponding to the interlaced scanning is not a base vector. First inverse orthogonal transform means for performing an inverse orthogonal transform on the block; and an orthogonal transform block of the compressed image data, which has been orthogonally transformed by an orthogonal transform method (frame orthogonal transform mode) corresponding to progressive scanning, is a base vector. If not, a second inverse orthogonal transform unit for performing an inverse orthogonal transform on the orthogonal transform block, and an orthogonal transform block of the compressed image data that has been orthogonally transformed in the field orthogonal transform mode is a base vector A third inverse orthogonal transform means for performing inverse orthogonal transform on the orthogonal transform block, and a frame orthogonal transform mode When the orthogonal transformation block of the compressed image data subjected to the orthogonal transformation by the transformation is a base vector, the orthogonal transformation block comprises a fourth inverse orthogonal transformation means for performing an inverse orthogonal transformation on the orthogonal transformation block. The inverse orthogonal transform means performs inverse orthogonal transform on a coefficient of a low frequency component among the coefficients of the orthogonal transform block, and the second inverse orthogonal transform means performs a coefficient of all frequency components of the orthogonal transform block. To each pixel of the orthogonally transformed block subjected to the inverse orthogonal transformation into two pixel blocks corresponding to the interlaced scanning, and orthogonally transformed each of the two separated pixel blocks. The inverse orthogonal transform is performed on the coefficients of the low frequency components among the coefficients of the two transformed pixel blocks, and the two orthogonally transformed pixel blocks are combined to generate an orthogonal transform block. The third inverse orthogonal transformation means is configured to perform an orthogonal transformation on coefficients of all frequency components of the orthogonal transformation block of the compressed image data which has been subjected to the orthogonal transformation in the field orthogonal transformation mode, with respect to a curve obtained by performing an inverse orthogonal transformation. Using the inverse orthogonal transform coefficient of the low-frequency component of the phase, performs the inverse orthogonal transform, and the fourth inverse orthogonal transform unit performs an orthogonal transform on the compressed image data orthogonally transformed in the frame orthogonal transform mode. The method according to claim 1, wherein the inverse orthogonal transform is performed on a curve obtained as a result of inverse orthogonal transform of coefficients of all frequency components, using an inverse orthogonal transform coefficient of a low frequency component having a phase on the same curve. The image decoding device according to any one of the preceding claims. 3. The first inverse orthogonal transform means performs an inverse orthogonal transform on a coefficient of a low frequency component among the coefficients of the orthogonal transform block, and performs a top orthogonal transform of the top field obtained by performing the inverse orthogonal transform. A phase correction of 1/4 pixel is performed in the vertical direction of each pixel, and a phase correction of 3/4 pixel is performed in the vertical direction of each pixel of the bottom field obtained by performing the inverse orthogonal transform. The second inverse orthogonal transform means performs inverse orthogonal transform on the coefficients of all the frequency components of the orthogonal transform block, and executes the orthogonal transform block on which the inverse orthogonal transform has been performed by interlaced scanning.
Into two pixel blocks, perform orthogonal transformation on the two separated pixel blocks, and perform inverse orthogonal transformation on the low-frequency component coefficients among the coefficients of the two orthogonally transformed pixel blocks. With respect to the vertical direction of each pixel of the bottom field obtained by performing the inverse orthogonal transformation, the phase is corrected by 1/4 pixel with respect to the vertical direction of each pixel of the top field obtained by performing the orthogonal transformation. The phase correction of 3/4 pixel is performed, and the top field and the bottom field whose phases have been corrected are combined. The third inverse orthogonal transform unit performs the orthogonal transform in the field orthogonal transform mode on the compressed image data. This is the inverse orthogonal transform coefficient of the low-frequency component on the same curve as the curve obtained as a result of inverse orthogonal transform of the coefficients of all the frequency components of the orthogonal transform block. Of the low-frequency component of the phase delayed by ４ pixel with respect to the vertical direction of each pixel of the bottom field obtained by performing the inverse orthogonal transform with a delay of １ ／ pixel in the vertical direction. The orthogonal transform coefficient is used to perform an inverse orthogonal transform. The fourth inverse orthogonal transform means performs inverse orthogonal transform on coefficients of all frequency components of the orthogonal transform block of the compressed image data that has been orthogonally transformed in the frame orthogonal transform mode. The inverse orthogonal transform coefficient of the low-frequency component on the same curve with respect to the curve obtained as a result of the conversion. The inverse orthogonal transform coefficient is delayed by 1/4 pixel with respect to the vertical direction of each pixel of the top field. 3 with respect to the vertical direction of each pixel of the bottom field
3. The image decoding apparatus according to claim 2, wherein the inverse orthogonal transform is performed using an inverse orthogonal transform coefficient of a low-frequency component having a phase delayed by / 4 pixels. 4. A predetermined pixel block (macro block)
From the compressed image data of the first resolution, which has been subjected to prediction coding by performing motion prediction in units and compression coding by performing orthogonal transformation in units of predetermined pixel blocks (orthogonal transformation blocks), In an image decoding method for decoding moving image data having a second resolution lower than the resolution, an inverse orthogonal transform is performed on the low frequency component coefficients among the orthogonal transform blocks of the orthogonally transformed compressed image data. Add the compressed image data subjected to the inverse orthogonal transform and the reference image data subjected to the motion compensation, and store the moving image data obtained by the addition as reference image data; and a macro of the stored reference image data. When motion compensation is performed on the block and the coefficients of the orthogonal transform block are base vectors, the results of inverse orthogonal transform of the coefficients of all frequency components of the orthogonal transform block are performed. Against the curve obtained by using the inverse orthogonal transform coefficients of the low frequency components of the phase on the same curve, the image decoding method characterized by the inverse orthogonal transform. 5. When the orthogonal transformation block of the compressed image data that has been subjected to the orthogonal transformation by the orthogonal transformation method (field orthogonal transformation mode) corresponding to the interlaced scanning is not a base vector, among the coefficients of the orthogonal transformation block, If the orthogonal transform block of the compressed image data obtained by performing the inverse orthogonal transform on the coefficients of the low frequency component and performing the orthogonal transform by the orthogonal transform method (frame orthogonal transform mode) corresponding to the progressive scanning is not a base vector, Inverse orthogonal transformation is performed on the coefficients of all the frequency components of the orthogonal transformation block, and each pixel of the orthogonal transformation block subjected to the inverse orthogonal transformation is separated into two pixel blocks corresponding to interlaced scanning,
The orthogonal transformation is performed on each of the two separated pixel blocks, and the inverse orthogonal transformation is performed on the coefficient of the low frequency component among the coefficients of the orthogonally transformed two pixel blocks, and the two pixels are subjected to the inverse orthogonal transformation. When the orthogonal transform block of the compressed image data subjected to the orthogonal transform in the field orthogonal transform mode is a base vector, the orthogonal transform block is generated by combining the blocks, and the orthogonal transform is performed in the field orthogonal transform mode. For the curve obtained as a result of inverse orthogonal transformation of the coefficients of all frequency components of the orthogonal transformation block of the compressed image data, perform inverse orthogonal transformation using the inverse orthogonal transformation coefficient of the low frequency component having the phase on the same curve, If the orthogonal transform block of the compressed image data subjected to the orthogonal transform in the frame orthogonal transform mode is a base vector, the frame orthogonal transform is performed. For the curve obtained by inverse orthogonally transforming the coefficients of all the frequency components of the orthogonally transformed block of the compressed image data that has been orthogonally transformed by the mode, the inverse orthogonal transform coefficient of the low frequency component having the phase on the same curve is used. And performing an inverse orthogonal transform.
3. The image decoding method according to claim 1. 6. When the orthogonal transform block of the compressed image data subjected to the orthogonal transform by the orthogonal transform method (field orthogonal transform mode) corresponding to the interlaced scanning is not a base vector, each of the coefficients of the orthogonal transform block is used. The inverse orthogonal transform is performed on the coefficients of the low-frequency component, the phase is corrected by 1/4 pixel in the vertical direction of each pixel of the top field obtained by the inverse orthogonal transform, and the inverse orthogonal transform is performed. The compressed image data obtained by performing a phase correction of 3/4 pixel with respect to the vertical direction of each pixel of the bottom field obtained by the above, and performing an orthogonal transformation by an orthogonal transformation method (frame orthogonal transformation mode) corresponding to progressive scanning. If the orthogonal transformation block is not a base vector, the orthogonal transformation block that performs inverse orthogonal transformation on the coefficients of all frequency components of the orthogonal transformation block and performs inverse orthogonal transformation It is separated into two pixel blocks corresponding to the interlaced scanning, and the two separated pixel blocks are subjected to orthogonal transformation. The coefficients of the two orthogonally transformed pixel blocks are inversely transformed with respect to the coefficient of the low frequency component. Each pixel of the bottom field obtained by performing a quadrature transform, performing a phase correction of 1/4 pixel in the vertical direction of each pixel of the top field obtained by performing the inverse orthogonal transform, and performing the inverse orthogonal transform. 3/4 of the vertical direction
If the orthogonally transformed block of the compressed image data subjected to the orthogonal transformation in the field orthogonal transformation mode is a base vector, the top field and the bottom field having undergone the phase compensation for the pixels are combined, and the phase compensation is performed. An inverse orthogonal transform coefficient of a low frequency component on the same curve as a curve obtained as a result of inverse orthogonal transform of coefficients of all frequency components of the orthogonal transform block of the compressed image data orthogonally transformed by the orthogonal transform mode. , With respect to the vertical direction of each pixel in the top field,
Using the inverse orthogonal transform coefficient of the low-frequency component of the phase delayed by 3/4 pixel with respect to the vertical direction of each pixel of the bottom field obtained by delaying 4 pixels and performing inverse orthogonal transform If the orthogonal transform block of the compressed image data subjected to the orthogonal transform in the frame orthogonal transform mode is a base vector, the orthogonal transform of the compressed image data subjected to the orthogonal transform in the frame orthogonal transform mode is performed. This is an inverse orthogonal transform coefficient on the same curve with respect to a curve obtained as a result of inverse orthogonal transform of the coefficients of all frequency components of the transform block, and is a delay of 1/4 pixel in the vertical direction of each pixel of the top field And the inverse orthogonal transform using the inverse orthogonal transform coefficient of the low-frequency component of the phase delayed by 3/4 pixel with respect to the vertical direction of each pixel of the bottom field obtained by performing the inverse orthogonal transform. The image decoding method according to claim 5, wherein: