JP2000023195A - Image encoding apparatus and method, image decoding apparatus and method, and encoded data providing medium - Google Patents

Abstract

(57) [Problem] In MPEG4, a pixel value per pixel defines an image having information of 4 to 12 bits in both luminance and chrominance as an input / output image.
When the coefficients are encoded in the same manner as MPEG2,
There is a case where inconvenience occurs due to insufficient calculation accuracy. SOLUTION: A quantization step generator 308 controls a DC component coefficient (D) of orthogonal transform data according to a desired image quality of an input image and a pixel precision indicating a pixel value per pixel.
C encoding) is determined. DC coefficient quantizer 3
07 quantizes the DC coefficient of the orthogonal transform data according to the encoding accuracy determined by the quantization step generator 308.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding method and apparatus for outputting an encoded bit stream obtained by encoding a moving image, for example, and an image decoding method and apparatus for decoding the encoded bit stream. The present invention relates to a providing medium for providing the coded bit stream.

[0002]

2. Description of the Related Art For example, in a system for transmitting moving image data to a remote place, such as a video conference system or a video telephone system, image data is converted into a line correlation or a frame in order to use a transmission path efficiently. The compression encoding is performed using the inter-correlation.

A moving picture experts group (MPEG) is a typical moving picture coding scheme.
(Moving picture coding for storage). This is ISO-I
Discussed in EC / JTC1 / SC2 / WG11,
It has been proposed as a standard proposal, and employs a hybrid method combining motion compensation prediction coding and DCT (Discrete Cosine Transform) coding.

[0004] In MPEG, several profiles and levels are defined in order to support various applications and functions. Profiles define the classification of functions, and levels define the differences in quantity. The most basic is the main profile main level (MP @ M
L (Main Profile at Main Level).

FIG. 70 is a diagram showing MP @ M in the MPEG system.
5 shows an exemplary configuration of an L encoder.

[0006] The input image (image data) to be encoded is
The data is input to the frame memory 31 and is temporarily stored. Then, the motion vector detector 32 reads out the image data stored in the frame memory 31 in units of macroblocks composed of, for example, 16 pixels × 16 pixels, and detects the motion vector.

In the motion vector detector 32, the image data of each frame is converted into one of an I picture (intra-frame coding), a P picture (forward prediction coding), and a B picture (bidirectional prediction coding). Process as It should be noted that whether the image of each frame that is sequentially input is to be processed as an I, P, or B picture is determined by, for example, I, B, P, B, P,.
Is predetermined.

That is, the motion vector detector 32 refers to a predetermined reference frame in the image data stored in the frame memory 31, and the reference frame and the current encoding target are By performing pattern matching (block matching) with a 16-pixel × 16-line macro block of a frame, a motion vector of the macro block is detected.

[0009] Here, in MPEG, the image prediction modes include four modes: intra coding (intra-frame coding), forward prediction coding, backward prediction coding, and bidirectional prediction coding.
There are types, I-pictures are intra-coded, P-pictures are coded by either intra-coding or forward-prediction coding, and B-pictures are intra-coded, forward-prediction-coded, backward-prediction-coded, or both. Encoded by any of the method predictive encodings.

That is, the motion vector detector 32 sets the intra coding mode as the prediction mode for the I picture. In this case, the motion vector detector 32
The motion vector is not detected, and the prediction mode (intra prediction mode) is output to the VLC (variable length coding) unit 36 and the motion compensator 42.

The motion vector detector 32 performs forward prediction on a P picture and detects the motion vector. Further, the motion vector detector 32 compares a prediction error caused by performing forward prediction with, for example, a variance of a macroblock to be coded (a macroblock of a P picture). As a result of the comparison, when the variance of the macroblock is smaller than the prediction error, the motion vector detector 32 sets the intra coding mode as the prediction mode, and outputs it to the VLC unit 36 and the motion compensator 42. If the prediction error caused by performing the forward prediction is smaller, the motion vector detector 32 sets the forward prediction encoding mode as the prediction mode, and sets the VLC unit 36 and the motion compensator 42 together with the detected motion vector.
Output to

Further, the motion vector detector 32 performs forward prediction, backward prediction, and bidirectional prediction on the B picture, and detects respective motion vectors. Then, the motion vector detector 32 performs forward prediction, backward prediction,
And a minimum prediction error of the bidirectional prediction (hereinafter, appropriately referred to as a minimum prediction error), and the minimum prediction error and the variance of the encoding target macroblock (the macroblock of the B picture), for example. Compare with As a result of the comparison, if the variance of the macroblock is smaller than the minimum prediction error, the motion vector detector 32 sets the intra coding mode as the prediction mode, and sets the VLC unit 3
6 and the motion compensator 42.

If the minimum prediction error is smaller, the motion vector detector 32 sets the prediction mode in which the minimum prediction error is obtained as the prediction mode, and sets the VLC unit 36 and the motion vector together with the corresponding motion vector. Compensator 42
Output to The motion compensator 42 includes the motion vector detector 3
When both the prediction mode and the motion vector are received from 2,
According to the prediction mode and the motion vector, the encoded and already locally decoded image data stored in the frame memory 41 is read, and the read image data is used as the prediction image data by the calculator 33.
And 40.

The arithmetic unit 33 reads from the frame memory 31 the same macroblock as the image data read from the frame memory 31 by the motion vector detector 32, and calculates the difference between the macroblock and the predicted image from the motion compensator 42. Is calculated. This difference value is supplied to the DCT unit 34.

On the other hand, when only the prediction mode is received from the motion vector detector 32, that is, when the prediction mode is the intra-coding mode, the motion compensator 42 does not output a predicted image. In this case, the operator 33 (the operator 40
Does not perform any particular processing, and converts the macroblock read from the frame memory 31 into the DCT unit 34 without any processing.
Output to

Here, the properties of the two-dimensional DCT will be described with reference to FIG. For example, when two-dimensional DCT is applied to a two-dimensional block consisting of 8 lines × 8 pixels, FIG.
As shown in FIG. 7, an 8 × 8 DCT coefficient F (x, y) is generated.
Among these coefficients, it is known that the coefficient F (0,0) in the 0th row and the 0th column corresponds to a DC component representing an average luminance value in a two-dimensional block.

F (1,0), F (2,0),..., F (6,0),
It is known that coefficients arranged in the right direction, such as F (7,0), represent high frequency components in the vertical direction in a two-dimensional block. This means that the following rows F (1,1), F (2,1), ..., F (6,
The same applies to 1) and F (7,1). On the other hand, F (0,1), F (0,
It is known that coefficients arranged in a downward direction, such as 2),..., F (0,6) and F (0,7), represent high-frequency components in the horizontal direction in a two-dimensional block. This is the same for the following columns.

The DCT coding is performed in the image 2 of the image signal.
Utilizing the dimensional correlation, signal power is concentrated on a specific frequency component, and only the concentratedly distributed coefficients are encoded, so that the amount of information can be compressed. For example, in a block having a flat picture and a high autocorrelation of an image signal (each pixel level in the block is almost equal), low frequency components (F (0,0), F (1,0), F (0 , 1), DCT coefficients near F (1, 1)) show large values, and the other coefficients are almost zero. Therefore, the amount of information can be compressed by using a technique of omitting continuous identical coefficients by Huffman coding or the like.

The DCT unit 34 performs the above-mentioned DCT processing consisting of 8 lines × 8 pixels on the output data of the arithmetic unit 33, and converts the resulting DCT coefficient into a quantizer 35.
To supply.

In the quantizer 35, of the DCT coefficients, the coefficient of the DC component is converted into a predetermined value (in MPEG2, any one of 8, 9, 10, and 11 is used to encode the image quality of the image to be encoded and the macroblock of the image). Linear quantization in the quantization step (selected according to the encoding method), and the fraction is rounded off. Similarly A
For the C coefficient, the quantizer 35 stores the amount of data stored in the buffer 37 (the amount of data stored in the buffer 37).
A quantization step of the AC component is set in accordance with (buffer feedback), and the DCT coefficient from the DCT unit is quantized in the quantization step. Quantized DC
The T coefficient is sent to a scan converter (scan converter) 43.

The quantized DCT coefficients are zigzag scanned by the scan converter 43 in the order of low to high frequency coefficients, and sent to the DC coefficient differentiator 44 as a one-dimensional signal.

In the DC coefficient differentiator 44, the DC component coefficients of the DCT coefficients are differentiated between adjacent blocks. Differentiation is performed in a different manner between the luminance (Y) block and the two color difference (Cb, Cr) blocks.

A method of differentiating the DC component of the DCT coefficient will be described with reference to FIG. That is, FIG. 72 (a) shows a method of differentiating a luminance block, and the DC component coefficients of each block are differentiated between the DC component coefficients of the upper, lower, left and right adjacent blocks in a zigzag order as shown in FIG. And stored again in each block. In the color difference block, FIG.
As shown in (b), the difference is made between the DC component coefficients of each of the blocks adjacent on the left and right, and is stored again in each block.

A DC coefficient differentiator 4 for performing these differences
73 and (a) and (b) of FIG. 73 show the specific configurations of the inverse differentiator corresponding thereto, respectively. However, the first block (the first intra-coded block after the inter-coded block or the first block of the slice) cannot be differentiated (neighboring blocks are not aligned) Therefore, a predetermined value is used as the initial value (128 in the case of MPEG2 when quantized by the quantizer 35, 256 in the case of 9 and 51 in the case of 10).
In the case of 2, 11, 1024) is given, and the difference from this initial value is obtained.

The quantized DCT coefficients (hereinafter, appropriately referred to as quantization coefficients) are supplied to the VLC unit 36 together with the set quantization steps.

The VLC unit 36 converts the quantized coefficient supplied from the quantizer 35 into a variable length code such as a Huffman code and outputs the converted code to a buffer 37. further,
The VLC unit 36 performs a quantization step from the quantizer 35,
The prediction mode (mode indicating which of intra coding (intra-picture predictive coding), forward predictive coding, backward predictive coding, and bidirectional predictive coding has been set) and motion from the motion vector detector 32 The vector is also variable-length coded, and the resulting coded data is
Output to

Here, the DCT in the VLC unit 36 is particularly used.
An encoding method and a decoding method of the DC component coefficient of the coefficient will be described with reference to Tables 1, 2, and 3.

[0028]

[Table 1]

[0029]

[Table 2]

[0030]

[Table 3]

First, a DC component coefficient (DIFFERN
From TIAL DC), refer to Table 1 for the size dct dc size
(= 0, 1,...) And calculate it in Table 2.
Alternatively, encoding is performed according to Table 3. That is, for example, when the DC component coefficient (DIFFERNTIAL DC) is +5, the size is set to 3 according to Table 1. And 3 as this size,
The right column of Table 2 or Table 3 is coded to the code 101 or 110 described in the left column of the third row.

Next, referring again to Table 1, a fixed-length code (Co) representing a differential DC component coefficient (DIFFERNTIAL DC) is referred to.
de) (fixed-length code representing a coefficient value having a bit width equal to the size) is obtained, and a differential DC component coefficient value is transmitted by a combination of these two codes.

The variable length code representing the size is different between the luminance (Y) block and the chrominance (Cb, Cr) block, and the encoding is performed by referring to Table 2 for the luminance block and to Table 3 for the chrominance block. Will be DC component coefficient (DIFF
The fixed length code (Code) having a bit width equal to the size representing the value of the ERNTIAL DC) is one-to-one with the coefficient value as shown in Table 1.
Is supported.

For example, as described above, the difference value is +5, and when the difference value is that of a luminance block, the size is 3 in Table 1 and the sign is 101 in Table 2. The fixed-length code representing the value +5 is 1 from Table 1.
01. Therefore, the code output for the difference value +5 is a 6-bit code 101101 obtained by combining these.

The buffer 37 in FIG. 70 temporarily stores the encoded data from the VLC unit 36 to smooth the data amount, and outputs the encoded data as an encoded bit stream, for example, to a transmission path, or to a recording medium. Record.

The buffer 37 outputs the data storage amount to the quantizer 35, and the quantizer 35 sets a quantization step according to the data storage amount from the buffer 37. That is, the quantizer 35 outputs the buffer 3
When 7 is about to overflow, the quantization step is increased, thereby reducing the data amount of the quantization coefficient. When the buffer 37 is about to underflow, the quantizer 35 reduces the quantization step, thereby increasing the data amount of the quantization coefficient. Thus, the overflow and the underflow of the buffer 37 are prevented.

The quantization coefficient and the quantization step output from the quantizer 35 are supplied not only to the VLC unit 36 but also to the inverse scan converter 45.

The inverse scan converter 45 performs an inverse zigzag scan of the quantized DCT coefficient from its low-frequency component to a high-frequency component, reconstructs it into an 8 × 8 matrix, and outputs it to the inverse quantization circuit 38.

The inverse quantizer 38 inversely quantizes the quantized coefficient from the quantizer 35 in accordance with the quantization step from the quantizer 35, thereby converting the quantized coefficient into a DCT coefficient. This DCT coefficient is calculated by an IDCT unit (an inverse DCT unit).
39. The IDCT unit 39 performs an inverse DCT process on the DCT coefficient, and supplies data obtained as a result of the process to the arithmetic unit 40. .

The arithmetic unit 40 is supplied with the same data as the predicted image supplied to the arithmetic unit 33 from the motion compensator 42, as described above, in addition to the output data of the IDCT unit 39. The arithmetic unit 40 locally decodes the original image data by adding the output data (prediction residual (difference data)) of the IDCT unit 39 and the predicted image data from the motion compensator 42, and locally decodes the original image data. The output image data (local decoded image data) is output (however, when the prediction mode is the intra coding, the output data of the IDCT unit 39 passes through the arithmetic unit 40 and is directly output to the locally decoded image data. The data is supplied to the frame memory 41 as data).
The decoded image data is the same as the decoded image data obtained on the receiving side.

The decoded image data (local decoded image data) obtained by the arithmetic unit 40 is supplied to and stored in the frame memory 41, and then inter-coded (forward predictive coding, backward predictive coding, quantity direction predictive coding). It is used as reference image data (reference frame) for the image to be encoded.

FIG. 74 shows an example of the configuration of an MPEG @ ML decoder for decoding encoded data output from the encoder shown in FIG.

A coded bit stream (coded data) transmitted via a transmission path is received by a receiver (not shown), or a coded bit stream (coded data) recorded on a recording medium is shown in FIG. The data is reproduced by the reproduction device, supplied to the buffer 101, and stored.

The IVLC unit (inverse variable length decoder) 102
Read the encoded data stored in the buffer 101,
By performing variable-length decoding, the encoded data is separated into a motion vector, a prediction mode, a quantization step, and a quantization coefficient for each macroblock. Among these data, the motion vector and the prediction mode are supplied to the motion compensator 107, and the quantization step and the quantization coefficient of the macroblock are supplied to the DC coefficient inverse difference unit.

In the DC coefficient inverse differentiator 108, the operation described with reference to FIG. 73B is performed, and the output is output to the inverse scan converter 109 together with the AC component of the DCT coefficient.

The inverse scan converter 109 operates in the same manner as the inverse scan converter 45 shown in FIG. 70, and its output is sent to the inverse quantizer 103.

The inverse quantizer 103 inversely quantizes the macroblock quantization coefficient supplied from the inverse scan converter 109 in accordance with the quantization step also supplied from the IVLC unit 102, and obtains the resulting DCT.
The coefficients are output to the IDCT unit 104. IDCT device 10
4 is the DCT of the macroblock from the inverse quantizer 103
The coefficients are subjected to inverse DCT and supplied to the arithmetic unit 105.

The arithmetic unit 105 is supplied with the output data of the motion compensator 107 in addition to the output data of the IDCT unit 104. That is, the motion compensator 107 converts the already decoded image data stored in the frame memory 106 into I
It is read out according to the motion mode and the prediction mode from the VLC unit 102, and the arithmetic unit 1
05. The arithmetic unit 105 outputs the output data (prediction residual (difference value)) of the IDCT unit 104 and the motion compensator 10
7, the original image data is decoded. The decoded image data is supplied to and stored in the frame memory 106. If the output data of the IDCT unit 104 is intra-coded, the output data passes through the arithmetic unit 105 and is supplied to the frame memory 106 as decoded image data and stored as it is. You.

The decoded image data stored in the frame memory 106 is used as reference image data of image data to be decoded thereafter. Further, the decoded image data is
As an output reproduction image, for example, it is supplied to a display (not shown) or the like and displayed.

In MPEG1 and MPEG-2, B pictures are not used as reference image data, and are not stored in the frame memory 41 in FIG. 70 or the frame memory 106 in FIG. 74 in each of the encoder and the decoder.

The encoder shown in FIG. 70 and FIG.
Is based on the MPEG1 / 2 standard. Currently, however, the decoder performs encoding in units of VO (Video Object), which is a sequence of objects such as objects constituting an image. ISO-IEC / JTC
In 1 / SC29 / WG11, MPEG (MovingPi
standardization work is being carried out as the “Cture Experts Group” 4.

[0052]

By the way, the above MPE
In G4, an image whose pixel value per pixel has information of 4 bits to 12 bits in both luminance and chrominance is defined as the input / output image.
When the method is performed in the same manner as in the case of PEG2, the calculation accuracy may be insufficient, which may cause inconvenience.

That is, in the case of the one-dimensional DCT processing, it is known that the value after the DCT processing is approximately 2√2 times the value before the DCT processing. Since the range of the pixel value of the input image is 8 bits (0 to 255), the range of the DC coefficient of the two-dimensional DCT transform coefficient is approximately 8 (= 2√2 × 2√2) times, that is, 11 bits. (0 to 2047).

In MPEG2, this 11-bit precision value is quantized with values of 8, 4, 2, 1 according to the image quality of the output image as shown in Table 4, and the precision is set to 8, It is possible to select from 9, 10, and 11.

[0055]

[Table 4]

MPEG4 does not currently define a method for changing the DCT calculation accuracy in accordance with the desired image quality. That is, since the minimum value of the quantization scale of the DC component of the DCT coefficient is defined as 8, the upper limit of the bit accuracy of the DC component is set to the number of bits of the pixels of the input image. For example, in the case of 10 bits, the bit precision of the DC component is fixed at 10 bits.

Also, when encoding the AC component of the DCT, the value of the AC component is set to [-2048: 2047] regardless of the number of bits per pixel of the input image. Therefore, when the number of bits per pixel is large, inconvenience may occur when this rule is used.

As described above, the encoding accuracy of the DC coefficient of the DCT is determined only by the image quality of the input image, and the encoding range of the AC component that can be encoded is fixed regardless of the number of bits per pixel. This means that when transmitting a higher-quality moving image using MPEG4, the image quality must be reduced from what was originally desired,
It becomes a problem.

However, regarding the method of calculating the DC coefficient value and the method of encoding the DCT coefficient used in MPEG2, the pixel value of the input image is specified only for 8 bits, and the other bit numbers are particularly specified. Not. Therefore, when a method similar to the encoding method of the DCT coefficient of MPEG2 is used, inconvenience may occur in the image encoding of MPEG4.

That is, when the encoding method of the DCT DC component similar to MPEG2 is applied in MPEG4, the maximum value of the encoding precision is 11 bits. It is conceivable that the accuracy is insufficient and the image quality deteriorates more than originally desired.

The present invention has been made in view of such circumstances, and changes in coding accuracy of DC coefficients of DCT are adaptively performed according to required image quality and pixel accuracy of an input image. It is another object of the present invention to provide an image encoding apparatus and method for realizing high-quality and high-efficiency encoding by applying an AC component encoding process according to the pixel accuracy of an input image. It is another object of the present invention to provide a decoding device and method for decoding an encoded bit stream by the above-described image encoding. It is still another object of the present invention to provide an encoded data providing medium for providing an encoded bit stream by the image encoding.

[0062]

On the encoding device side, according to the required image quality and the number of bits of one pixel of the input image,
The required precision (number of bits) of the DC component coefficient of the orthogonal transform such as DCT is transmitted in units of VOP, GOV, or video object layer, and the encoder adapts according to the information and the number of bits per pixel. To change the coding method of the DC component coefficient, further expand the variable length code table for the DC component coefficient, and output the variable length code.Also, regarding the coding method of the AC component, It is characterized in that the encoding means for the DC and AC components is adaptively changed according to the pixel accuracy around one pixel.

On the decoding device side, the transmitted D
Accuracy of DC component coefficient of orthogonal transform such as CT (number of bits)
In the video object layer or GOV or video object plane, and the decoder adaptively decodes the DC component coefficient according to the information and the bit amount per pixel transmitted in the video object layer. Of the variable length code table for DC component coefficients, and
Decoding DC component coefficients of orthogonal transform such as CT, and also regarding the encoding method of AC component, adaptively changing the decoding accuracy based on the pixel accuracy of one pixel of the input image, and DC, AC for pixel accuracy
The encoding means of the component is adaptively changed.

More specifically, in an image encoding apparatus that encodes an image and outputs a resulting encoded bit stream, the image encoding apparatus performs a function according to a desired image quality for the image and a pixel precision indicating a pixel value per pixel. Encoding accuracy determining means for determining the encoding accuracy of the DC component coefficients of the orthogonal transform data, and quantizing the DC component coefficients of the orthogonal transform data according to the encoding accuracy determined by the encoding accuracy determining means And DC component coefficient quantization means.

Also, in an image encoding method for encoding an image and outputting an encoded bit stream, the orthogonal transform data is converted according to a desired image quality for the image and a pixel precision indicating a pixel value per pixel. The encoding accuracy of the DC component coefficient is determined, and the DC component coefficient of the orthogonal transform data is quantized according to the determined encoding accuracy.

Further, in an image coding apparatus for coding an image and outputting a coded bit stream obtained as a result, a main part of the orthogonal transform data is determined in accordance with a pixel precision indicating a pixel value per pixel of the image. And a coding unit for coding each of the residual and the residual part separately.

Also, in an image encoding method for encoding an image and outputting an encoded bit stream obtained as a result, the orthogonal transform data is converted in accordance with the pixel precision indicating the pixel value per pixel of the image. It is characterized in that the main part and the residual part are separately encoded.

Further, in an image decoding apparatus for decoding an image signal which has been subjected to intra- and inter-image encoding processing and pixel data subjected to orthogonal transformation and quantization processing, a DC component of the orthogonal transformation data obtained by the orthogonal transformation is provided. Inverse quantization means for inversely quantizing the coefficients, and a decoding accuracy used in the inverse quantization processing of the inverse quantization means adaptively in accordance with a signal representing the precision of the DC component coefficient and the pixel precision of the image. And a decoding accuracy determining means for determining.

Further, in an image decoding method for decoding an image signal subjected to intra-image and inter-image encoding processing and pixel data subjected to orthogonal transformation and quantization processing, the orthogonal transformation data of the orthogonal transformation is used. The decoding accuracy when the DC component coefficient is inversely quantized is adaptively determined according to the signal representing the accuracy of the DC component coefficient and the pixel accuracy of the image.

In an image decoding apparatus for decoding an image signal which has been subjected to intra-image and inter-image encoding processing, pixel data subjected to orthogonal transformation, and quantization processing, the pixel value represents one pixel value of an image. And decoding means for separately decoding the main part and the residual part of the quantized orthogonally transformed data depending on the pixel precision used for the decoding.

Further, in an image decoding method for decoding an image signal which has been subjected to intra-image and inter-image encoding processing and pixel data subjected to orthogonal transformation and quantization processing, a pixel value per one pixel of an image is obtained. The main part and the residual part of the quantized orthogonal transform data are separately decoded according to the pixel precision used to represent.

Further, the encoding accuracy of the DC component coefficient of the orthogonal transform data is determined by the encoded data providing medium according to the desired image quality of the image and the pixel accuracy indicating the pixel value per pixel. Provided is an encoded bit stream obtained by quantizing the DC component coefficients of the orthogonal transform data according to the encoding accuracy.

Further, an encoded bit stream obtained by separately encoding the main part and the residual part of the orthogonally transformed data according to the pixel precision indicating the pixel value per pixel of the image is provided. .

According to the encoding apparatus and method of the present invention, the required accuracy (bits) of the DC component coefficients of the orthogonal transform such as DCT according to the required image quality and the amount of bits per pixel. ), The encoding method is adaptively changed, the variable length code table is extended, and the variable length code is output, so that efficient encoding is possible.

Further, according to the decoding apparatus and method of the present invention, according to the accuracy of the DC component coefficient of orthogonal transform such as DCT transmitted and the bit amount per pixel of the output image,
Since the decoding method is adaptively changed, and the table of the variable length code is extended and decoded, the decoding can be performed without waste. Therefore, in the conventional MPEG2, the accuracy of the DC component coefficient corresponds to only 8 bits per pixel of the output image, but the number of bits per pixel can be applied to an arbitrary image.

Further, it corresponds to a coding method of an AC component of DCT having a pixel precision of 8 bits or more. For example, when coding an input image having a pixel precision of 8 bits or more, a coding method which is not wasted is selected. It is possible to

[0077]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Some embodiments of the present invention will be described below. First, in the first embodiment, M
The encoding precision is adaptively changed based on the DCT encoding and decoding methods used in PEG4. The present invention can be applied not only to MPEG4 but also to other moving picture signal transmission systems. That is,
In each of the following embodiments, a case will be described where the color difference format of the input moving image signal is a 4: 2: 0 format, but the accuracy is 4: 2: 0.
The present invention can be applied to cases other than the format.

First, before describing the first embodiment of the present invention, Visu as a term necessary for understanding the present invention is described.
al Object, video Object, Video Object Layer (VO
L), Group of Video Object Plane (GOV), Vide
o Object Plane (VOP), Group of block (GO
B), macroblocks (MB), and blocks will be briefly described with reference to FIG.

A block is composed of, for example, 8 lines × 8 pixels adjacent to each other for each luminance or color difference. For example, DCT is performed in this unit. FIG. 2 shows the arrangement of blocks in the MB in the 4: 2: 0 format.

For example, when the image format is a so-called 4: 2: 0 component digital signal, an MB (macroblock) is composed of four luminance blocks Y0, Y1, Y2, and Y3 that are vertically and horizontally adjacent to each other. In this example, each of the color difference blocks of Cb and Cr at the same position is composed of a total of six blocks. The order of transmission is Y0, Y1, Y2, Y3, C
b, Cr. What to use for the prediction data, whether or not it is not necessary to send the prediction error, and the like are determined in this unit.

A VOP is a single image composed of one or a plurality of N macroblocks MB. Then, according to the encoding method, the image is classified into one of an I picture (intra-coded image), a P picture (forward predicted coded image), and a B picture (bidirectional predicted coded image).

In the IVOP, the image itself is encoded (intra-encoded) without performing motion compensation. PVO
P is basically forward predictive coded based on an image (I or PVOP) located earlier in time than itself. A B picture is basically bidirectionally coded based on two pictures (I or PVOP) located before and after the B picture.

The GOV includes at least one IVOP,
0 or at least one non-IVOP (PVOP, BV
OP). However, this does not apply to the upper layer when hierarchical coding is performed. In addition, GO
The use or non-use of V is determined freely during encoding.

A VOL (Video Object Layer) is composed of at least one VOP or GOV. V
The OL does not include VOPs having the same display time,
By displaying OPs in the order of their display time, a series of images is obtained.

A Video Object is composed of one or more VOLs, and the same object can be encoded and decoded at a plurality of resolutions, frame rates, and the like by combining VOLs.

A Visual Object is created by one Video Object or another Object (for example, face object, m
esh object, still texture object, etc.).

The Visual Object Sequence is composed of one or a plurality of Visual Objects.

Next, an encoder serving as an application example of the image encoding method and apparatus according to the first embodiment of the present invention will be described with reference to FIG.

Image data to be encoded (input image signal)
Is input to a VO (Video Object) composing unit 1. In the VO composing unit 1, a VO, which is a sequence thereof, is constructed for each object constituting an image input thereto.
OP component is output 2 ₁ to 2 _N. That is, when N pieces of VO # 1 to VO # N are configured in the VO constructing unit 1, the N pieces of VO # 1 to VO # N are output to the VOP configuring unit 2 ₁ to 2 _N.

Specifically, for example, when the image data to be encoded is composed of a sequence of independent background F1 and a sequence of foreground F2, the VO constructing unit 1 converts the sequence of foreground F2 to VO As # 1, VOP
And outputs to the configuration unit 2 _1, the sequence of the background F1, as VO # 2, and outputs the VOP constructing unit 2 _2.

When the image data to be encoded is, for example, the background F1 and the foreground F2 already synthesized, the VO constructing unit 1 divides the image into regions according to a predetermined algorithm. Background F1 and foreground F
2 and VO as each sequence
To the corresponding VOP constituent unit 2 _n (where n = 1, 2,.
.., N).

The VOP constructing unit 2 _n constructs a VOP (Video Object Plane) from the output of the VO constructing unit 1. That is, for example, an object is extracted from each frame, and, for example, a minimum rectangle (hereinafter, appropriately referred to as a minimum rectangle) surrounding the object is set as a VOP. At this time, the VOP configuring unit 2 _n configures the VOP such that the number of horizontal and vertical pixels is a multiple of, for example, 16. VO component 2
_{When n} constructs a VOP, it outputs the VOP to the VOP encoding unit 3n.

Further, the VOP constructing unit 2 _n , size data (VOP size) indicating the size of the VOP (for example, the horizontal and vertical lengths) and the position of the VOP in the frame (for example, detecting the offset data (VOP offset) indicating the coordinates) at which the origin, even these data, and supplies the VOP encoding unit 3 _n.

The VOP encoder 3 _n outputs the output of the VOP constructor 2 _n to, for example, MPEG or H.264. In accordance with a control signal input from the outside in a system conforming to a standard such as H.263, a bit stream obtained as a result is
Output to The multiplexing unit 4 includes VOP coding units 3 _{1 to} 3 _n
, And the resulting multiplexed data (bitstream) is transmitted via, for example, a terrestrial wave, a satellite line, a CATV network, or another transmission path 5,
Alternatively, for example, the information is recorded on a recording medium 6 such as a magnetic disk, a magneto-optical disk, an optical disk, a magnetic tape, or the like.

Here, VO is a sequence of each object (object) constituting the composite image when a sequence of the composite image exists, and VOP means VO at a certain time. That is, for example, if there is a composite image F3 composed by combining the images F1 and F2, an image in which the images F1 or F2 are arranged in time series is
Each is a VO, and the image F1 or F2 at a certain time is a VOP. Therefore, a VO can be regarded as a set of VOPs of the same object at different times.

For example, when the image F1 is set as the background and the image F2 is set as the foreground, the composite image F3 is formed by using the key signals for extracting the image F2 and the images F1 and F2.
2, the VOP of the image F2 in this case includes not only the image data (luminance signal and color difference signal) constituting the image F2 but also its key signal as appropriate.

The sequence of image frames (image frames) does not change in both size and position, but VO
The size and position may change. That is, the same VO
May be different in size and position depending on the time.

A specific description will be given with reference to FIG. FIG. 4 shows a composite image including an image F1 as a background and an image F2 as a foreground.

The image F1 is, for example, a photograph of a certain natural scene, and the sequence of the entire image is one VO (VO # 0). Also, the image F2
Is an image of a person walking, for example, and the smallest rectangular sequence surrounding the person is one V
O (referred to as VO # 1).

In this case, since VO # 0 is a landscape image, basically, both its position and size do not change, similarly to a normal image frame. On the contrary,
Since VO # 1 is an image of a person, the size and position of the VO # 1 change when the person moves left and right or moves forward or backward in the drawing. Therefore, FIG.
Represents VO # 0 and VO # 1 at the same time, but the position and size of VO may change over time.

Therefore, the VOP encoding unit 3 _n in FIG. 3 outputs, in the output bit stream, information on the position (coordinate) and size of the VOP in a predetermined absolute coordinate system in addition to the data obtained by encoding the VOP. Has been made to include. In FIG. 4, the vector indicating the position of the VOP (image F1) at a certain time, which constitutes VO # 0, is represented by O.
ST0 and V of VO # 1 at the same time
OST1 is a vector representing the position of OP (image F2);
Each is represented.

[0102] Next, referring to FIG. 5 for detailed configuration of the VOP encoding unit 3 _n of FIG. FIG.
70, parts corresponding to those in FIG. 70 are denoted by the same reference numerals. That is, the VOP encoding unit 3 _n
Basically, the configuration is the same as that of the encoder shown in FIG. The VOP encoding unit 3 _n whose detailed configuration is shown in FIG. 5 can control the encoding method by inputting the control signal N. As a method of controlling the encoding means, for example, it represents the encoding accuracy of the DC component of the DCT. "
There are "intra_dc_precision", "number_of_bits" representing the pixel precision per pixel of the input image, and "fixed_intradc_q" which controls the DCT DC component encoding means.

"Intra_dc_precision", "numbe"
The r_of_bits "and" fixed_intra_q "codes are
5, is supplied to the inverse quantization circuit 138, the DC coefficient differentiator 144, and the VLC unit 136. The details of each will be described below.

The image data of the input image is supplied to and stored in the frame memory 31 as in the case of FIG. 70, and the motion vector detector 32 detects a motion vector in macroblock units.

That is, as described above, since the size and position of the VOP change depending on the time (frame),
In detecting the motion vector, it is necessary to set a coordinate system serving as a reference for the detection and to detect a motion in the coordinate system. Therefore, here, the motion vector detector 32 uses the above-described absolute coordinate system as a reference coordinate system, and uses the absolute coordinate system as a reference in accordance with the size data FSZ_B and the offset data FPOS_B. Place a VOP to
A motion vector is detected.

The detected motion vector (MV)
Is supplied to the VLC unit 136 and the motion compensator 42 together with the prediction mode.

Also, when performing motion compensation, it is necessary to detect motion in the reference coordinate system, as described above. Therefore, the size data FSZ_B and the offset data FPOS_
B is supplied.

The VOP in which the motion vector is detected is shown in FIG.
0, the VLC unit 1
36. FIG. 70 also shows the VLC device 136.
As in the case of, the quantization coefficient, the quantization step,
In addition to the motion vector and the prediction mode, size data FSZ_B and offset data FPOS_B from the image layering unit are also supplied, where all of these data are variable-length coded.

The VOP in which the motion vector is detected is coded as described above, and also locally decoded as in FIG. 70, and stored in the frame memory 41.

The arithmetic unit 33 reads the same macroblock as the image data read from the frame memory 31 by the motion vector detector 32 from the frame memory 31 in the same manner as in FIG. The difference from the prediction image is calculated. This difference value is DC
It is supplied to the T unit 34.

On the other hand, the motion compensator 42, as in FIG.
When the prediction mode is the intra-coding mode, no predicted image is output. In this case, the operator 33 (the operator 4
0 is the same), the frame memory 31
The macro block read out from the
4 is output.

The DCT unit 34 performs a DCT process consisting of 8 lines × 8 pixels on the output data of the arithmetic unit 33, and supplies a DCT coefficient obtained as a result to the quantizer 135.

At this time, the bit precision of the output DCT coefficient is represented by n + 3 bits in the case of an input with n-bit precision.

At the quantizer 135, the input DC
The T coefficient is quantized in the same manner as the quantizer 35 in FIG. 70 and sent to the scan converter 43.

However, the DC component of the DCT coefficient is quantized by the following method. Here, switching the coding precision (the number of bits) of the DC coefficient (DC coefficient) among the DCT coefficients in units of VOP, GOV, and VOL according to the required image quality and the pixel precision of the input image. It will be described in detail.

The coding precision of the DC coefficient is specified based on the "intra_dc_precision" code input as a control signal from the outside and the pixel precision "number_of_bits" per pixel. The encoding precision (number of bits) of VOP, GOV, V
When switching is performed in OL units, the data is described and transmitted in a VOP header, a GOV header, and a VOL header. “intra_dc_precision” is, for example, a 2-bit code, and can represent four types of encoding precision (bit precision). Also, pixel accuracy "number"
“_of_bits” is, for example, a 4-bit flag, and its value indicates pixel precision. “number_of_bits” is VOL
Transmitted by Hereinafter, "intra_dc_precision", for example, "00", "01", "10" or "11" and "n
In the case where the bit length per pixel indicated by “umber_of_bits” is n bits, it is assumed that encoding is performed with the bit precision of the DC coefficient as n, n + 1, n + 2 or n + 3 bits, respectively.

That is, if “number_of_bits” is 8, D
Bit precision of C coefficient is 8, 9, 10, 11 bits and intr
It changes according to the value of a_dc_precision. Similarly, in the case of 12 bits, the values are 12, 13, 14, and 15. Table 5 shows the relationship between the change in the bit precision of the DC coefficient, number_of_bits, and the quantization number of the DC coefficient.

[0118]

[Table 5]

The quantization circuit 135 in FIG. 5 is configured, for example, as shown in FIG. 6, in which the DCT coefficient output from the DCT unit 34, the prediction mode, the quantization step,
And "intra_dc_precision" and "fixed_intradc_q" are supplied.

In the prediction mode, the intra-flag generator 31
1 is input. The intra flag generator 311 sets the intra flag only when the prediction mode is the “intra coding (intra-picture prediction) mode” (the intra flag that is normally 0 is set to 1).

On the other hand, the DCT coefficients of the 8 × 8 block, that is, 64 DCT coefficients are supplied to the switch 300. The switch 300 selects the terminal A when the intra flag is 0, and selects the terminal B when the intra flag is 1.

Therefore, when the motion compensation mode is not the “intra coding (intra-picture prediction) mode”, the DC
The T coefficient is supplied to the quantizer 304 via the terminal A by the switch 300. A quantization step is input to the quantizer 304, where the DCT coefficients are quantized according to the quantization step. The quantized DCT coefficients are output to the blocking circuit as quantized coefficients.

The quantization coefficient is converted into 8
The image data is divided into blocks of x8 and output to the scan converter 43 in FIG. 5 as block quantization coefficients. In the case of MPEG, the quantizer 304 usually discards fractions after the decimal point during quantization.

On the other hand, when the motion compensation mode is the “intra coding (intra-picture prediction) mode”, the DCT coefficient
The coefficient is supplied to a coefficient separator 306. The DC / AC coefficient separator 306 separates the DCT coefficient into an AC coefficient and a DC coefficient, and outputs the AC coefficient to the quantizer 305 and the DC coefficient to the DC coefficient quantizer 307, respectively.

In the quantizer 305, as in the case of the quantizer 304, according to the quantization step,
The AC coefficients are quantized and output to the blocking circuit 309 as AC quantized coefficients.

The DC coefficient is linearly quantized (however, the fractional part is rounded off) in the DC coefficient quantizing circuit 307 in accordance with the quantization step generated by the quantization step generator 308, and the block is converted as a DC quantized coefficient. Output to the conversion circuit 309.

Here, the quantization step generator 308
Indicated DC coefficient encoding accuracy (bit accuracy) "intra
_dc_precision "and a flag" fixed_intradc_q "representing a quantization step determination method, a quantization step is generated.

The operation of the quantization step generator 308 will be described with reference to the flowchart of FIG. The quantization step generator 308 is "intra_dc_precision", "fixed_intrad
c_q "as its input.

In S1 of FIG. 7, when the flag "fixed_intradc_q" indicating the method of determining the quantization step is "0", the method of determining the quantization step moves to S2. S2
, The value of the quantization step of the DC coefficient is determined according to the following Table 6 from the value of the input quantization scale.

[0130]

[Table 6]

That is, if the block that is performing quantization is a luminance block and the input quantization scale is 1 to 4, the DC quantization step is 8, and similarly, the quantization scale is 5 to 8 In the case of, the value of 2 × quantization scale is used, and in the case of 9 to 24, the quantization scale is +8, 25
In the above case, the quantization scale is 2-16. Similarly, in the case of the color sub-block, the value of the DC quantization step is determined from the quantization scale and Table 6 above.

If the value of "fixed_intradc_q" is "1" in S1 of FIG. 7, the process moves to S3. In S3, "intra_d
The operation moves to S4 and S5 according to the value of "c_precision". When "intra_dc_precision" is "00", S is performed.
It moves to 4 and outputs 8 as a DC quantization step.
If "intra_dc_precision" is other than "00" in S3, S
Go to 5.

At S5, the operation moves to S6 and S7 according to the value of "intra_dc_precision". "intra_dc_pr
If "ecision" is "01", the flow shifts to S6, and outputs 4 as a DC quantization step.In S5, "intra_dc_precis"
If “ion” is other than “01”, the process moves to S7.

In S7, the operation moves to S8 and S9 according to the value of "intra_dc_precision". "intra_dc_pr
If "ecision" is "10", the flow shifts to S8, where 2 is output as the DC quantization step. In S7, "intra_dc_precis"
If “ion” is other than “10”, that is, “11”, the flow shifts to S9, where 1 is output as the value of the DC quantization step.

Therefore, when the motion compensation mode is the “intra coding (intra-picture prediction) mode”, the DC coefficient is quantized to the designated bit precision by the DC coefficient quantizer 307 in FIG. Will be.

The AC quantized by the quantizer 305 in FIG.
The coefficients and the DC coefficients quantized by the DC coefficient quantizer 307 are divided into 8 × 8 blocks by a blocking circuit 309 and output to the scan converter 43 in FIG. 5 as block quantized coefficients.

As described above, the quantization circuit 135 performs the quantization processing according to the required image quality and the bit precision of the input image.

Returning to FIG. 5, the scan converter 43 supplied with the quantized DCT coefficient output from the quantization circuit 135 operates in the same manner as that shown in FIG. The output from the scan converter 43 is sent to the DCT coefficient differencer 144 and the inverse scan converter 45.

The DC coefficient differentiator 144 is constructed, for example, as shown in FIG. 8, and includes a quantized DCT coefficient, a quantization scale, and pixel accuracy of one pixel “number_of_bits” and “intra_”.
dc_precision "as its input and performs the DC coefficient difference.

The prediction mode is determined by the intra flag generator 404.
Is input to The intra-flag generator 404 performs the same operation as the intra-flag generator 311 in FIG. 6, and raises the intra-flag only when the prediction mode is the “intra-coding (intra-picture prediction) mode” (normally 0). Is set to 1).

On the other hand, the quantized DCT coefficients supplied from the scan converter 43 of FIG.
The T coefficient is supplied to the switch 403. Switch 40
3 selects the terminal A side when the intra flag is 0,
When the intra flag is 1, the terminal B is selected.

Therefore, when the motion compensation mode is not the “intra coding (intra-picture prediction) mode”, the DC
The T coefficient is output from the DC coefficient differentiator 144 via the terminal A of the switch 403, and is output from the VLC unit 136 in FIG.
Supplied to.

On the other hand, when the motion compensation mode is “intra coding”, the DCT coefficients are supplied to DC / AC coefficient separator 402. DC / AC coefficient separator 40
2 operates in the same manner as the DC / AC coefficient separator 306 in FIG.
The DC component is supplied to the arithmetic unit 410 and the DC coefficient inverse quantizer 406.

The input quantization step, intra_dc
_precision, fixed_intradc_q is supplied to a quantization step generator 409, which operates in the same manner as the quantization step generator 308 in FIG. Quantizer 40
6 and the DC coefficient quantizer 405.

The DC coefficient inverse quantizer 406 performs the same operation as the DC coefficient inverse quantizer 315 in FIG. 9 showing the inside of the inverse quantizer 138 in FIG.
The C coefficient is dequantized by a DC coefficient quantization step, and the value is supplied to a DC coefficient prediction mode accumulation buffer 408. Note that the inverse quantizer 138 in FIG.
Will be described later.

Returning to FIG. 8, the DC coefficient / prediction mode accumulation buffer 408 receives the output from the intra flag generator 404 and the output 406 of the DC coefficient dequantizer and receives the position of the block, the state of the intra flag, Accumulate the inversely quantized DC coefficient.

The DC predicted value selector 407 outputs the predicted value of the DC coefficient using the DC coefficient and the information stored in the prediction mode accumulation buffer 408, and supplies the value to the DC coefficient quantizer 405.

Here, FIG. 10 is used to explain the operation of the DC prediction value selector 407. FIG. 10 shows a positional relationship between a block X for performing DC coefficient differentiation and a block for referring to a DC value in performing a difference. Here, the block on the left side of the block X is A, the block on the upper side is C, and the block on the upper left side is B.

Here, the DC component of block A is represented by F _A [0]
[0], the DC component of block B is F _B [0] [0], and the DC component of block C is
Let the C component be F _C [0] [0]. These values are the values of the DC components that have been inversely quantized, and the DC components of the quantized DC components of the A, B, and C blocks are obtained using the “intra_dc_precision” and the quantization scale according to the flowchart in FIG. It is obtained by integrating the values of the quantization steps. However, when these blocks are located outside the VOP or are blocks other than the intra block, the value of the power of 2 (number_of_bits + 2) from Table 6 above is used.

Here, the block used to calculate the DC difference value of the block X is determined by the following equation (1).

If (| F _A [0] [0] -F _B [0] [0] | <| F _B [0] [0] -F _C [0] [0] |) predict from block C else predict from block A ... (1) That is, the absolute value of the difference between F _A [0] [0] and F _B [0] [0] and F _B [0] [0]
And the absolute value of the difference between F _C [0] [0] and F _A [0] [0] and F _B [0] [0]
If the absolute value of the difference between F _B [0] [0] and F _C [0] [0] is smaller, block A is used as the prediction block and F _A [0] [0] is used as the prediction. The absolute value of the difference between F _B [0] [0] and F _C [0] [0] is F
_A [0] [0] and F _B [0] and the prediction value F _C [0] [0] and the block C and the prediction block is smaller than the absolute value of the difference between [0].

The predicted value of the DC coefficient thus obtained is supplied to the DC coefficient quantizer 405 in FIG.

In the DC coefficient quantizer 405, the DC coefficient quantizer 405 shown in FIG.
The same operation as the coefficient quantizer 307 is performed, and the supplied DC
The prediction value of the DC coefficient input in the coefficient quantization step is quantized and output to the differentiator 410. In the differentiator 410, the quantized D supplied from the DC coefficient separator 402
The difference between the C coefficient and the predicted value after quantization supplied from the DC coefficient quantizer 405 is obtained, and the difference is input to the DC / AC coefficient combiner 401.

As described above, the DC coefficient differentiator 144 shown in FIG.
The DC coefficient difference processing is performed according to the encoding accuracy (the number of bits) of the DC coefficient which is switched in units, GOV units, and VOL units.

The DC / AC coefficient combiner 401 combines the input DC coefficient difference value and the AC coefficient, and outputs the result as 64 DCT coefficients. This value is output to the VLC unit 136 in FIG.

The VLC coupler 136 in FIG.
OS_B, VOP_SIZE, VOP_offset, motion vector, prediction mode, quantization scale, quantized DCT coefficients, and the like are supplied, and each is VLC-coded and output to the buffer 37.

Here, a method of encoding the quantized DCT coefficients will be described in detail. The coding of the DC coefficient obtained by quantizing and calculating the difference is performed as follows. here,
Tables 7, 8, 9, and 10 are used in calculating the difference value of the DC coefficient.

[0158]

[Table 7]

[0159]

[Table 8]

[0160]

[Table 9]

[0161]

[Table 10]

First, the DC component coefficient (DIFFERN
TIAL DC), referring to Table 9 or Table 10, SIZE (= 0,
,..., 15), and encodes them according to Table 7 or Table 8. That is, for example, the DC component coefficient (D
If IFFERNTIAL DC) is +5, the size is set to 3 from Table 10. Then, 3 as this size is encoded to the code 010 or 001 described in the left column of the row of the third row in the right column of Table 7 or Table 8.

Next, referring again to Table 10, a fixed-length code (Co) representing a differential DC component coefficient (DIFFERNTIAL DC) will be described.
de) (fixed-length code representing a coefficient value having a bit width equal to the size) is obtained, and a differential DC component coefficient value is transmitted by a combination of these two codes.

The variable length code representing the size is different between the luminance (Y) block and the chrominance (Cb, Cr) block, and the encoding is performed by referring to Table 7 for the luminance block and Table 8 for the chrominance block. Will be DC component coefficient (DIFF
The fixed length code (Code) having a bit width equal to the size representing the value of (ERNTIAL DC) has a one-to-one correspondence with the coefficient value as shown in Table 6 above.

For example, in the same manner as described above, the difference value is +5, and when the difference value is that of a luminance block, the size is 3 in Table 10 and the sign is 010 in Table 7. Also, the fixed-length code representing the value of +5 is 101 from Table 10. Thus, for the difference value +5, the output code is a 6-bit code 0
10101.

The following procedure is used for encoding the AC coefficients. That is, the data supplied to the VLC unit 136 in FIG. 5 is the DCT coefficients arranged one-dimensionally by the scan converter 43, and the DC coefficients are further differentiated by the DC coefficient differentiator 144. is there. Therefore,
In the VLC unit 136, it is represented by the number of continuous zero coefficients (RUN) and the non-zero coefficient value (Level).

That is, D supplied to the VLC unit 136
CT coefficient is, for example, 12 (DC coefficient) 0 0 0 6 0 0 0 0 0 -2 0 0 3 0 0 0
In the case of the following expression, since the first 12 are DC coefficients, they are coded by the above-described DC coefficient coding method.
After the C coefficient, there are three 0 coefficients, so RUN is 3, and the non-zero coefficient is 6, so the Level is 6. Similarly, the next RUN is 5, Level is -2, and the last coefficient is RUN2 Leve
It is represented by l3. Thus, the 63 AC coefficients are represented by one or more pairs of RUN and Level.

Furthermore, in addition to the combination of RUN and Level,
There is a value LAST indicating whether the last is a non-zero AC coefficient. This LAST is if the non-zero coefficient indicated by Level is the last non-zero coefficient in the block (in scan order),
It becomes 1 in other cases, and becomes 0 otherwise.

For example, the AC coefficient after the 3
If the coefficient values are all 0, the AC coefficient becomes RUN 3 Level 6 Last 0 RUN 5 Level -2 Last 0 RUN 2 Level 3 Last 1, and this method can express all 63 AC coefficients. Become.

The AC coefficient converted into (RUN, Level, Last) in this manner is encoded by one of the following two patterns.

One is VLC shown in Tables 11 to 22 below.
This is a method of encoding from a table.

[0172]

[Table 11]

[0173]

[Table 12]

[0174]

[Table 13]

[0175]

[Table 14]

[0176]

[Table 15]

[0177]

[Table 16]

[0178]

[Table 17]

[0179]

[Table 18]

[0180]

[Table 19]

[0181]

[Table 20]

[0182]

[Table 21]

[0183]

[Table 22]

The codes of the combinations having a high frequency of occurrence are determined in advance in Tables 11 to 22, and the codes are output by referring to this table. Table 11
Table 16 is a VLC table in the intra macroblock, and Tables 17 to 22 are VLC tables other than the intra macroblock. Table 11 to Table 16
And s expressed in the column of VLC Code in the tables of Tables 17 to 22 represent the sign of Level, and is 0 when Level is a positive number and 1 when Level is a negative number.

Escape codes are used for VLC tables not shown in Tables 11 to 16 and 17 to 22. The escape code is a code written in the last line of Table 16 and Table 22, and when this flag is detected, RUN, Level, Last is obtained by the following means. That is, the encoding / decoding method is determined based on the bit following the escape code, and there are three types of encoding methods when 0 bit is input and 1 and 11 are input.

First, when a bit indicating a value of 0 is encoded / decoded following the escape code, RUN, Level is determined by using the VLC tables shown in Tables 11 to 16 and 17 to 22.
Are encoded / decoded. However, the Level represented by this code is different from the actual value, and the absolute value of the Level is corrected by the following procedure. This modification is performed as shown in equation (2), where LMAX is obtained by VLC encoding / decoding according to Tables 23 and 24 for intra blocks and Tables 25 and 26 for non-intra blocks. RUN, Last

LEVEL ^S = sign (LEVEL ⁺ )-abs (LEVEL ⁺ ) + LMAX] (2)

[0188]

[Table 23]

[0189]

[Table 24]

[0190]

[Table 25]

[0191]

[Table 26]

Next, when "10" is obtained according to the escape code, the same as in the method described above, Tables 11 to 16,
Run, Level using VLC tables shown in Tables 17 to 22
l is encoded / decoded. Where R
UN is different from the actual value.
The absolute value of is corrected. This modification is performed as shown in equation (3), where LMAX is calculated by VLC encoding / decoding according to Tables 27 and 28 for intra blocks and Tables 29 and 30 for non-intra blocks. It is obtained by using the obtained RUN and Last.

RUN ^S = RUN ⁺ + (RMAX + 1) (3)

[0194]

[Table 27]

[0195]

[Table 28]

[0196]

[Table 29]

[0197]

[Table 30]

When "11" is obtained following the escape code, Run, Last, and Level are all encoded / decoded with fixed bits. That is, RUN is 6 bits, Level is 15
The bit LAST is encoded / decoded with one bit.

Here, returning to FIG.
The output of 3 is input not only to the DC coefficient differentiator 144 but also to the inverse scan converter 45 which is complementary to the scan converter 43. Then, in the inverse scan converter 45, inverse zigzag scanning is performed in the order of low-frequency to high-frequency coefficients, and converted into a signal. The signal is inversely quantized by an inverse quantizer 138 complementary to the quantizer 135 based on a quantization step (inverse quantization step), and a DCT coefficient is output.

Here, the inverse quantizer 138 will be described with reference to FIG. The inverse quantizer 138 is configured as shown in FIG. 9 and includes a quantized DCT coefficient, a motion compensation mode, a quantization step (inverse quantization step), and “i”.
"ntra_dc_precision" code and "fixed_intradc_q" are supplied.

The motion compensation mode is input to the intra flag generator 311. The intra flag generator 311
The same operation as the intra flag generator 311 in FIG.
Only when the motion compensation mode is the “intra coding (intra-picture prediction) mode”, an intra flag is set (usually 0
Is set to 1).

On the other hand, the quantized D of an 8 × 8 block
The CT coefficients, ie, the 64 quantized DCT coefficients, are provided to switch 300. The switch 300 selects the terminal A when the intra flag is 0, and selects the terminal B when the intra flag is 1. Therefore, the motion compensation mode is “intra coding (intra-picture) Otherwise, the quantized DCT coefficient is supplied to the inverse quantizer 313 via the terminal A by the switch 300. The inverse quantizer 313 includes:
A quantization step (inverse quantization step) is input, and according to the inverse quantization step, the quantized DCT coefficients are inversely quantized and output to the blocker 309 as DCT coefficients.

The DCT coefficient is calculated by the blocker 309 into 8 ×
8 and are output to the inverse DCT circuit 39 in FIG. 5 as block DCT coefficients.

On the other hand, when the motion compensation mode is the “intra coding (intra-picture prediction) mode”, the quantized DCT coefficient is supplied to the DC / AC coefficient separator 312 by the switch 300 via the terminal B. Supplied. DC /
In the AC coefficient separator 312, the quantized DCT
The coefficients are separated into an AC coefficient and a DC coefficient, and the AC coefficient is supplied to the inverse quantizer 314 and the DC coefficient is supplied to the DC coefficient inverse quantizer 3.
15 respectively.

In the inverse quantizer 314, the inverse quantizer 315
As in the case of, the quantized AC coefficient is inversely quantized according to the inverse quantization step, and the AC coefficient is output to the blocker 309.

The quantized DC coefficient is linearly inversely quantized in the DC coefficient inverse quantizer 315 according to the quantization step generated by the quantization step generator 308, and is output as a DC coefficient to the block generator 309. Is done.

Here, the quantization step generator 308
As with the quantization step generator 308 shown in FIG.
Generate a quantization step corresponding to the "ntra_dc_precision" code.

Therefore, when the motion compensation mode is the “intra coding (intra-picture prediction) mode”, the DC coefficient quantized by the DC coefficient inverse quantizer 315 is based on the designated bit precision. It will be inversely quantized.

The AC coefficient from the inverse quantizer 314 and the DC
The DC coefficient from the coefficient inverse quantizer 315 is divided into 8 × 8 blocks by the block generator 309, and output to the inverse DCT circuit 39 in FIG. 5 as block DCT coefficients.

The output DCT coefficient is subjected to inverse DCT processing by an inverse DCT circuit 39 having a complementary relationship with the DCT unit 34, and is output as a difference signal.

In the adder 40 shown in FIG.
The predicted image output from the motion compensator 42 based on the motion compensation mode is added to the difference signal from the pixel-by-pixel unit, and decoded into image data similar to the original image data.
The locally decoded image data is written as an image to be used for forward / backward / bidirectional prediction at an address of a frame memory group 41 specified by an image instruction signal to be described later, which is output from the frame memory 41.

The image stored in the frame memory group 41 is output from the motion compensator 42 as an image used for backward prediction or an image used for forward prediction.

[0213] On the other hand, the motion compensator 42 applies the motion from the motion vector detector 32 to the image (locally decoded image) specified by the motion compensation reference image instruction signal stored in the frame memory group 41. The motion compensation is performed based on the compensation mode and the motion vector, a predicted image is generated, and the predicted image is output to the subtractor 33 and the adder 40. That is, only in the forward / backward / bidirectional prediction mode, the motion compensator 42 sets the read address of the frame memory group 41 from the position corresponding to the position of the block that the frame memory group 41 is currently outputting to the subtractor 33. The image data to be used for forward prediction or backward prediction is read out, shifted by an amount corresponding to the motion vector, and output as predicted image data.

In the bidirectional prediction mode, both images used for forward prediction and backward prediction are read, and, for example, the average value is output as predicted image data. The prediction image data is supplied to the subtractor 33, and difference data is generated as described above.

Further, the prediction image data is also supplied to the adder 40. In the case of forward / backward / bidirectional prediction, the adder 40 receives, from the inverse DCT circuit 39, the difference data obtained by differentiating the prediction image, in addition to the prediction image data. Is added to the predicted image from the motion compensator 42 to perform local decoding. This locally decoded image is exactly the same image as the image decoded by the decoding device. As described above, the frame memory is used as an image used when performing forward / backward / bidirectional prediction on the next processed image. Stored in group 41.

In the case of the intra prediction mode, the adder 4
Since the image data itself is sent to 0 as the output of the inverse DCT unit 39, the adder 40 outputs this image data as it is to the frame memory group 41 and stores it.

In MPEG4, MPEG1
And MPEG2, B picture (B-VOP)
Is also used as a reference image, so that the B picture is also locally decoded and stored in the frame memory 41. However, at present, the B picture is used as a reference image only in an upper layer when hierarchical coding is performed.

On the other hand, as described with reference to FIG. 70, the VLC unit 136 shown in FIG. 5 outputs I, P, and B pictures (I-VOP, P
-VOP, B-VOP) is determined as to whether or not to be a skipped macroblock, and flags COD and MODB indicating the determination result are set. The flags COD and MODB are also transmitted after being variable-length coded.

Next, a decoder which is an application example of the image decoding method and apparatus according to the first embodiment of the present invention will be described with reference to FIG. This decoder is shown in FIG.
The bit stream output from the encoder shown in (1) is decoded.

The decoder is supplied with a bit stream provided from the encoder shown in FIG. 3 via the transmission line 5 or the recording medium 6. That is, the bit stream output from the encoder of FIG. 3 and transmitted via the transmission path 5 is received by a receiving device (not shown), or the bit stream recorded on the recording medium 6 is reproduced by a reproducing device (not shown). The data is reproduced by the device and supplied to the demultiplexer 71.

The demultiplexer 71 receives the bit stream (VS (Video Stream) described later) input thereto. Further, the demultiplexer 71 converts the input bit stream into a bit stream VO # for each VO.
, VO # 2,... And supplied to the corresponding VOP decoding unit 72 _n . The VOP decoding unit 72 _n converts the bit stream from the demultiplexing unit 71 into a VO
(Image data) and size data (VOP
size) and offset data (VOP offset), and supplies the decoded data to the image reconstruction unit 73.

[0222] In the image reconstruction unit 73, the VOP decoding unit 72
_The original image is reconstructed based on the output from each of _{1 to} 72 _N. The reconstructed image is supplied to the monitor 74 and displayed, for example.

FIG. 1 shows a specific configuration example of the VOP decoding section 72 _n .
It is shown in FIG. Note that, in the figure, parts corresponding to those in the decoder in FIG. 74 are denoted by the same reference numerals.
That is, the VOP decoding unit 72 _n is basically configured similarly to the decoder in FIG.

The lower layer bit stream from the demultiplexer 71 is supplied to the buffer 101, where it is received and temporarily stored. The IVLC unit 102 appropriately reads out a bit stream from the buffer 101 according to the processing state of the subsequent block, and performs variable-length decoding on the bit stream to obtain a quantization coefficient, a motion vector, a prediction mode, a quantization step, , Size data FSZ
_B, offset data FPOS_B, and flag C
Separate OD etc. The quantization coefficient and the quantization step are supplied to the inverse quantizer 103, and the motion vector and the prediction mode are determined by the motion compensator 107, the inverse quantizer 103,
It is supplied to a DC coefficient inverse differencer. Also, size data F
The SZ_B and the offset data FPOS_B are supplied to the motion compensator 107 and the image reconstruction unit 73 in FIG. "Intra_dc_precision", "number_of_bits", "fi
“xed_intradc_q” is supplied to the inverse quantizer 103, the DCT coefficient inverse difference unit 109, and the IDCT unit 104.

[0225] The inverse quantizer 103, IDCT unit 104, arithmetic unit 105, frame memory 106, the inverse scan converter 108 or the motion compensator 107, the inverse quantizer shown in FIG. 5 of the VOP encoding unit 2 _n 138, IDCT
The same processing as in the case of the unit 39, the arithmetic unit 40, the frame memory 41, the inverse scan converter 45, or the motion compensator 42 is performed.

In the inverse quantizer 103, the IVLC unit 102
Inverse quantization is performed based on the supplied quantization coefficient and the quantized DCT coefficient. The operation of the inverse quantizer 103 is the same as the operation of the inverse quantizer 38 in FIG. 5, and the output is sent to the inverse scan converter 108.

In the inverse scan converter 108,
The same operation as that of the inverse scan converter 45 of FIG. 5 is performed, and the DCT coefficients arranged one-dimensionally are inversely zigzag scanned and rearranged as 8 × 8 DCT coefficients. The output of the inverse scan converter 108 is sent to the DC coefficient inverse difference unit 109.

In the DC coefficient inverse differentiator 109, D in FIG.
The same operation as that of the C coefficient inverse differentiator 144 is performed, and the differentiated DC coefficient is decoded here. The result of the DC coefficient inverse differentiator 109 is supplied to the IDCT unit 104.

Here, a specific example of the DC coefficient inverse differentiator 109 will be described with reference to FIG. In FIG. 8, FIG.
The same numbers are assigned to those corresponding to the DC coefficient quantizers of No.

That is, the DC / AC coefficient combiner 401, DC
/ AC coefficient separator 402, switch 403, intra-flag generator 404, quantizer 405, inverse quantizer 40
6, the DC prediction value selector 407, the DC coefficient prediction mode accumulation buffer 408, and the quantization step generator 409 perform the same processing as that shown in FIG. The DC / AC coefficient combiner 401 further includes a DC value obtained from the DC / AC coefficient separator 402 and a DC predicted value selector 407.
Adder 4 with the quantized value of the DC prediction value obtained from
1 and the AC value obtained from the DC / AC coefficient separator 402 are combined and output.

As described above, the DC coefficient inverse difference unit 109
In the above, a DC coefficient is predicted by selecting a predicted value of the DC coefficient from the previously input DCT coefficient, and the sum of the DC components of the input DCT coefficient is decoded, and the DC coefficient is decoded and the value is converted to an IDCT unit
4 is output.

The IDCT device 104 is the IDCT device 39 shown in FIG.
The same operation as described above is performed. Here, the IDCT device 104
Performs an IDCT process on the DCT coefficient supplied from the DC coefficient inverse difference unit 109, and outputs the result to the arithmetic unit 105.

Arithmetic unit 105 performs the same operation as arithmetic unit 40 in FIG. 5, adds the predicted image supplied from motion compensator 107 and the decoded image supplied from IDCT unit 104, and stores the result in a frame memory. And then to the image reconstruction unit 73 of FIG. When performing intra coding (intra-frame coding), the motion compensator 10
7, the image supplied from the IDCT device 104 is output as it is.

The frame memory 106 operates in the same manner as the frame memory 41 shown in FIG. 5, accumulates the decoded image supplied from the arithmetic unit 105, and uses it as a predicted image when the motion compensator 107 at the subsequent stage performs motion compensation. Supply.

The motion compensator 107 operates in the same manner as the motion compensator 42 shown in FIG. 5, and outputs the motion vector, prediction mode, FSZ_B, FPOS_B, VOP supplied from the IVLC unit 102.
Based on _SIZE, VOP_Offset, an image to be used for predictive encoding is selected from within the frame memory and output to the arithmetic unit 105.

As described above, the VOP decoding unit 7 shown in FIG.
At 2 _n , the VOP is decoded and supplied to the image reconstruction unit 73.

Next, the syntax of the encoded bit stream output from the encoder shown in FIG. 3 will be described by taking, for example, Visual Comitee Draft (hereinafter, appropriately referred to as CD) of the MPEG4 standard as an example. In the first embodiment, the flag indicating the encoding accuracy of the DC coefficient of the intra macro block is arranged and described in the VOP layer.
This flag is VOL, VO, GOV, MB, bloc
It can also be represented by other layers such as k and GOB.

With respect to visual objects in MPEG4, not only two-dimensional moving image data but also two-dimensional still image images, face objects, and the like can be encoded / decoded. Therefore, as for the visual object of MPEG4, first, the Visual Object Sequence shown in FIG. 14 is transmitted. Within the Visual Object Sequence, multiple Visu
al Object can be transmitted.
Can configure a Visual Object Sequence.

Next, the syntax of the Visual Object is shown in FIG. 15 and FIG. In Visual Object, the type of object following this syntax (Video Object which is a moving image at present, Still Texture Obj which is a still image)
ect, Mesh Object showing 2D mesh, showing face shape
Face Object is defined), Video Object, Stil following Visual Object.
l The syntax of texture object, MeshObject, or Face Object follows.

[0240] Also, user_data shown in FIG.
ect Sequence, Visual Object, VideoObject, GOV layer, etc., and it is possible to define and transmit data used at the time of decoding at the time of encoding on the decoding side.

FIG. 18 shows the syntax of Video Object (hereinafter referred to as VO). A VO is composed of at least one VOL (Vi
deo Object Layer Class). However, when the image is not hierarchized, it is composed of one VOL. When the image is hierarchized, it is composed of VOLs of the number of layers.

Next, the syntax of the VOL (video Object Layer) is shown in FIGS. 19, 20, 21 and 22. The VOL is a class for scalability as described above, and is identified by a number indicated by video_object_layer_id. That is, for example, the lower layer V
The video_object_layer_id for the OL is set to 0, and, for example, the video_obje for the VOL of the upper layer
ct_layer_id is set to 1. As described above, the number of scalable layers is not limited to two, but may be any number including one or three or more.

[0243] The VOL is roughly composed of two syntaxes, and the video_object_layer_sta in FIG.
The part consisting of rt_code and short_video_start in FIG.
It consists of two parts starting with _marker.

Here, video_object_start_c in FIG.
The syntax consisting of the part following ode will be described. The flag indicating the bit precision of one pixel of the decoded image in each VOP is indicated by a 4-bit flag of bits_per_pixel. This bits_per_pixel is read only when the 1-bit flag not_8bit in FIG. 20 indicating whether the pixel precision of the input image determined for each VOL is 8 bits is 1. The value indicated by bits_per_pixel indicates the pixel accuracy at which decoding is performed, and the value is allowed to take a value from 4 to 12. When not_8_bits is “0”,
The pixel precision of the image to be decoded is processed as 8 bits.

Also, at the beginning of the VOL bit stream
short_video, not video_object_layer_start_code
When _start_marker is read, els at the bottom of FIG. 21
The syntax below the e-th row is used when decoding the VOL bit stream.

VOL is a single or a plurality of VOPs, vide
o_plane_with_short_header or GOV.

Now, the GOV layer will be described. GOV
The VOL layer and the VOP layer can be arbitrarily inserted into the coded bitstream at positions satisfying the conditions.
Specified between layers. As a result, a certain VOL #
0 is VOP # 0, VOP # 1,..., VOP # n,
VO such as VOP # (n + 1),..., VOP # m
In the case where the GOV layer is composed of a sequence of P, the GOV layer includes not only the immediately preceding VOP # 0 but also the VOP #
It can be inserted immediately before (n + 1). Therefore,
In the encoder, the GOV layer can be inserted, for example, at a position in the coded stream at which random access is desired. Therefore, by inserting the GOV layer, a series of VOPs constituting a certain VOL becomes: GO
Depending on the V layer, it is divided into a plurality of groups (hereinafter, appropriately referred to as GOV) and encoded. However,
It is stipulated that an IVOP comes in bit stream order immediately after a GOV header indicating a GOV. GOV (Gr
oup_of_Video Object Plane) layer syntax is defined, for example, as shown in FIG.

FIGS. 24, 25, 26, 27 and 2
8, FIG. 29, FIG. 30, FIG. 31, and FIG.
Shows the syntax of Plane. In FIG. 25, fixed_
intradc_q is D of DCT in Intra macro block.
This is a flag indicating a quantization method of the C coefficient. fixed_intrad
When the value of c_q is “0”, the values of Table 6
, The quantization step of the DC coefficient of the Intra block is determined.

At this time, since the quantization step of the DC coefficient is determined by the method shown in Table 6, the quantization scale of the DC coefficient changes non-linearly in the VOP according to the change of the value of the quantization scale. The value of fixed_intradc_q is "
In the case of "1", a 2-bit flag "intra_dc_precision" indicating the quantization precision of the DC coefficient of the intra macroblock is read in. If the value of fixed_intradc_q is 1, VO is read.
The set value of the quantization step of the DC value of the DCT coefficient in P is fixed. The value of the quantization scale is determined by the method described above, and the intra_dc_precision is “00”.
8 for “1”, 4 for “01”, 2 for “10”, “1”
In the case of 1 ", it becomes 1. The bit precision after quantization of the DC value at this time is a value shown in Table 5.

In this embodiment, fixed_intradc_q and i
The ntra_dc_precision has been described in the VOP, but the method of determining these flags and the quantization scale of the DC value is V
Not only OP, VO, VOL, GOV, MB, GO
B, block, etc. It is also possible to perform in other layers,
The setting is not limited to the VOP.

[0251] In video_object_plane, to encode texture information and shape encoding information of an image,
The tion_shape_texture is read, and in this, encoding such as macroblocks is performed.

The video_packet_header shown in FIG.
In the VOL, it can be used only when error_resilient_disable indicates “0”, and its use can be freely used on the encoding side and can be read from the VOP.

Here, returning to FIG. 21, the VOL is short_vi
vide used when starting with deo_start_marker
o_plane_with_short_header will be described. video_p
lane_with_short_header is VOL short as described above
Used only when starting with t_video_start_marker. FIG. 31 shows a configuration example of this syntax.

Short_video_start_marker is composed of a flag group and a plurality of gob_layers shown in FIG.

The gob_layer is obtained by coding a plurality of macroblocks as a group, and
The number of macroblocks in yer is uniquely determined by the picture frame of the image being encoded. Gob_layer also
It is determined by the syntax shown in FIG.

Referring back to FIGS. 24 to 27, video_object_pla
In ne, motion_shape_texture is read in order to encode texture information and shape encoding information of an image, and macroblocks and the like are encoded therein.

FIG. 33 shows motion_shape_texture. mo
tion_shape_texture is composed of two main parts, data_pattitioning_motion_shape_texture and comb
Divided into ined_motion_shape_texture. data_partit
The ioning_motion_shape_texture is used when the 1-bit flag data_partitioned shown in FIG. 20 of the VOL is 1, and the texture information is transmitted.

[0258] combined_motion_shape_texture is data_
This is used when partitioning is 0 or when only shape information is transmitted, and FIG. 34 shows the syntax. Thus, combined_motion_shape_texture is composed of one or a plurality of macroblocks.

FIGS. 35, 36 and 37 show examples of the syntax of a macroblock. The macro block syntax consists of three major parts, I and PV.
It is composed of three parts: a part indicating the syntax of the macroblock in the OP, a part indicating the syntax of the macroblock in the BVOP, and a part indicating the macroblock in the GrayScale shape.

I, a part indicating the syntax of a macroblock in PVOP and a part indicating the syntax of a MB in BVOP are a part m for encoding a block of shape.
b_binary_shape_coding, a group of flags indicating the MB encoding state, a motion vector encoding unit motion_vector, and each block encoding unit block. In addition, the portion indicating the GrayScale information of the macroblock is the Gray-Scale in the MB.
It is composed of a flag group indicating a state and an encoding unit alpha_block of a block configuring the Gray-Scale.

The mb_binary_shape_coding is composed of, for example, the syntax shown in FIG. Also, motion_vec
tor is composed of, for example, the syntax shown in FIG.

The block has, for example, the syntax shown in FIG. In the block, the difference value of the DC coefficient and its value are encoded by the method described above.
Subsequent DCT coefficients indicates other AC coefficients, which are also encoded by the above-described method.

Returning to FIGS. 24 to 32, data_partition
When the flag of ing is 1 and the texture information is transmitted, data_partitioning_motion_sha shown in FIG.
pe_texture is encoded. data_partitioning_motion
The _shape_texture is roughly composed of two parts, data_partitioning_I_VOP shown in FIGS. 42 and 43 and data_partitioning_I_VOP shown in FIGS. 44, 45 and 46.
It is composed of P_VOP.

Data_partitioning_I_VOP, data_partitio
Both the ning_P_VOP encodes a flag group indicating the nature of the VOP and the DCT coefficient of each block. The DCT coefficient encoding method is the same as the DCT coefficient encoding method described above.

Next, a second embodiment of the present invention will be described. In the second embodiment, “intra_dc_
The definition of the flag of "precision" is changed from that of the first embodiment. That is, "intra_dc_precision" is a 2-bit flag, and when its value is "00", the DC coefficient of the DCT coefficient is divided by 8. When the bit precision of one pixel of the input image is n bits, the bit precision of the quantized DC coefficient is defined as n bits.
In the case of "01", the DC coefficient is divided by 4 and the bit precision after quantization is n + 1 bits, and in the case of "10", the DC coefficient is divided by 2 and represented by n + 2 bits.
When “ra_dc_precision” is “11”, unlike the first embodiment, the encoder / decoder determines that the flag “fixed_intradc_q” indicating the coding method of the DC coefficient is 1, and the DC coefficient Is divided by the value determined according to Table 6. In the second embodiment, the flag “fixed_intradc_q” indicating the coding method of the DC coefficient of the intra block is not coded, and “intra_dc
_precision "determines the method of quantization and inverse quantization of the DC coefficient.

A specific example of the image signal encoding method and apparatus according to the second embodiment is the encoder shown in FIG. 3, similar to the specific example of the first embodiment, and has the same operation. I do. However, when the value of intra_dc_precision is “11”, the quantization step of the DC coefficient is determined by the method shown in Table 6.
The DC quantization step generator 308 shown in FIGS. 6 and 9 operates according to the flowchart shown in FIG. 47, and outputs a DC coefficient quantization step according to the operation.

In S11 of FIG. 47, “intra_dc_prec
When the value of "ision" is "11", the value of the quantization step of the DC coefficient is determined from the value of the input quantization scale in accordance with Table 6 above, that is, the block in which quantization is performed has the luminance If the block is a block and the input quantization scale is 1 to 4, the DC quantization step is 8
Similarly, when the quantization scale is 5 to 8, the value becomes 2 × quantization scale, when it is 9 to 24, the quantization scale + 8, and when it is 25 or more, the quantization scale × 2−16.
Similarly, in the case of the color difference block, the value of the DC quantization step is determined based on the quantization scale and Table 6.

If the value of “intra_dc_precision” is other than “11” in S11, the process moves to S13. S13
Then, according to the value of "intra_dc_precision", the operation is S1
4 and move to S15. "intra_dc_precision" is ”
In the case of "00", the process moves to S14 and outputs 8 as the DC quantization step. In S13, "intra_dc_precision" is changed to "
If it is not "00", the process moves to S15.

At S15, the operation moves to S16 and S17 according to the value of "intra_dc_precision". "intra
If "_dc_precision" is "01", the process moves to S16 and D
4 is output as a C quantization step. Returning to S15, if "intra_dc_precision" is "10", the process moves to S17. In S17, 2 is output as a DC quantization step.

In the encoder according to the second embodiment, the flag fi indicating the quantization method of the intra DC coefficient is used.
xed_intradc_q does not need to be used.

The encoder in the second embodiment is the same as that in the first embodiment shown in FIG. 5 and operates in the same manner. However, V shown in FIG.
In the VLC decoder in the OP decoder, intra_dc
When the value of _precision is “11”, the quantization step of the DC coefficient is determined by the method shown in Table 6. D and D shown in FIGS.
The C quantization step generator 409 operates according to the flowchart shown in FIG. 47, and outputs a DC coefficient quantization step according to the operation.

In the decoder according to the second embodiment, a flag fi indicating a method of quantizing intra DC coefficients is used.
xed_intradc_q does not need to be used.

The syntax in the second embodiment is the first except for the syntax of VideoObjectPlane.
This is the same as the embodiment. The syntax in the Video ObjectPlane is shown in FIG. 48, FIG. 49, FIG. 50 and FIG. Video Object Plane similar to the first embodiment
Other syntaxes have the same effect.

In the Video Object Plane syntax shown in FIGS. 48, 49, 50 and 51, “intra_dc_precision” in FIG. 48 is a flag indicating the quantization accuracy of the DC coefficient, Is determined by the method described above, and intra_dc_precision is “0”.
8 for “0”, 4 for “01”, 2, for “10”,
Becomes In the case of “11”, since the value is determined by the method shown in Table 6, the quantization scale of the DC coefficient changes nonlinearly in the VOP according to the change of the value of the quantization scale.

In the second embodiment, intra_dc_pre
Although the cision has been described with reference to the VOP, these flags and the method of determining the quantization scale of the DC value can be performed not only for the VOP but also for other layers such as VO, VOL, GOV, MB, GOB, and blocks. It is possible, and the setting is not limited to the VOP.

[0276] In video_object_plane, in order to encode texture information and shape encoding information of an image, mo
The tion_shape_texture is read, and in this, encoding such as macroblocks is performed.

Next, a third embodiment of the present invention will be described. In the third embodiment, the same processing as that of the first embodiment is performed, but the syntax below the output VOL is different only in the following output means. Here, a method of outputting a bit stream in the third embodiment will be described with reference to FIG.

In the third embodiment, a slice layer is adopted. The slice layer adopts the same definition as that defined in MPEG2.

[0279] A slice is composed of one or a plurality of MBs connected in the scanning order of an image. At the beginning of the slice, the difference between the motion vector and the DC component of the DCT coefficient in the image is reset, and the first MB has data indicating the position in the image, so that it can be restored even if an error occurs. Have been. Therefore, the length and the starting position of the slice can be arbitrarily changed depending on, for example, an error state of the transmission line. Also, the slice must end at the rightmost macroblock of the image and cannot cross a plurality of macroblock rows. That is, all the leftmost macroblocks of an image are always the start block of a slice, and the rightmost block is always an end macroblock. The start and end of the slice at other points are freely determined at the time of encoding.

An example of the structure of a slice in one frame is shown in FIG.
Shown in

In the third embodiment, D
The prediction of the C coefficient is always performed from the macroblock located on the left side. That is, in the first embodiment, the prediction value of the DC coefficient is predicted from the block located above or to the left of the macroblock to be coded as shown in FIG. But the third
In the embodiment, it is limited that the prediction is always performed from the macroblock located on the left side. If the first macroblock of the slice, that is, the left macroblock belongs to another slice, the DC prediction value is the first macroblock.
Is reset by the same method as that of the embodiment. The MV is similarly reset.

[0282] In the header of the slice, a Vertical MB Positi indicating the position of the head macroblock of the slice is provided.
on, Horizontal MB Position (see Figure 53), quantizer
Scale information is included, followed by a bit stream of one or more macroblocks.

[0283] Vertical MB Position, Horizontal MB Pos
ion gives the position of the MB in the vertical and horizontal directions at the beginning of the slice as described above. Vertical MB Posi
The “tion” and “Horizontal MB Position” are expressed in macroblock units, and their values indicate the position of the corresponding macroblock from the upper left of the screen with macroblock unit (16 pixel) accuracy. Also, quantizer_scale_code is used when calculating the quantization parameter in the first block of the slice.
Used as its predicted value.

Now, a method of outputting a bit stream in the encoder according to the third embodiment will be described with reference to FIG.

In S30 of FIG. 52, the encoder performs visu
l Output a bit stream according to the syntax of the Object sequence. The syntax of the Visual Object Sequence is the same as that shown in FIG. 14 of the first embodiment.

[0286] Next, proceeding to S31, the encoder performs visul Obje
Output a bitstream according to the syntax of ct. The syntax of the Visual Object is the same as that shown in FIGS. 15 and 16 of the first embodiment.

[0287] Next, proceeding to S32, the encoder performs the Video Obje
Output a bitstream according to the syntax of ct. The syntax of the Video Object is the same as that shown in FIG. 18 of the first embodiment.

Then, the flow advances to S33. Here, the video shown in FIGS. 54, 55, 56, 57, 58, and 59 is used.
Refer to Object Layer syntax. That is, the flag “fixed” indicating the quantization method in the intra block
_intradc_q ", the value of" fixed_intradc_q "is 0,
In the above-described first embodiment, when the quantization step of the DC coefficient in the intra block is determined by the method shown in Table 6, the process proceeds to S34 and the above Vi is performed.
Video_Ob in deo Object Layer syntax in Figure 54
A bit stream starting from ject_layer_start_code is output. Then, the process proceeds to S35, where Video Obje in FIG.
The output of the bit stream of ct Plane or less is performed.

Returning to S33, “fi_fi” is set by the flag “fixed_intradc_q” indicating the quantization method in the intra block.
If the value of “xed_intradc_q” is 1, that is, in the first embodiment described above, when the quantization step of the DC coefficient in the intra block is determined by the value of intra_dc_precision, the process proceeds to S36, and the Video_Object_layer_hig in FIG.
A bit stream starting from hquality_start_code is output. Then, the process proceeds to S37, where Video Object in FIG.
Output below Plane_highquality_header is performed.

As described above, the syntax for adaptively performing encoding is changed according to the value of “fixed_intradc_q”. In the third embodiment, "fix
Although the method of changing the syntax has been described using “ed_intradc_q”, the DC coefficient quantization method or the form of the syntax for performing encoding may be selected using other means.

Here, in the third embodiment,
Here is the syntax of the bitstream again.

The syntax of the Visual Object Sequence is the same as that of the first embodiment, and is shown in FIG. The syntax of the Visual Object is the same as that of the first embodiment, and is shown in FIGS. The syntax of the Video Object is the same as that of the first embodiment, and is shown in FIG.

The syntax of the Video Object Layer is shown in FIGS. 54, 55, 56, 57, 58 and 59. The syntax of the Video Object Layer is roughly divided into two parts. One is "fixed_intradc_
This is a case where the quantization step method of the DC coefficient is determined by using “Table 6” by “q”, and further, the quantization step is determined by using “intra_dc_precision” in “fixed_intradc_q”.

In summary, when “fixed_intradc_q” is 0, video_object_layer_start_c shown in FIG.
ode and short_video_start_marker shown in FIG.
Are encoded according to the syntax.

When “fixed_intradc_q” is 1, FIG.
Video_object_layer_highquality_s shown in 7
Encoded according to the syntax following tart_code.

Video_object_layer_highquality_start_c
When the bit stream is coded according to the syntax following ode, then, FIG. 60, FIG. 61 and FIG.
Is encoded following the video_object_plane_highquality_header shown in FIG. video_object_plane_hig
hquality_header is V as shown in FIGS. 63, 64 and 65.
It is a part corresponding to the ideo Object Plane and shows the same operation. However, some flag groups shown in the VideoObject Plane are not coded, and the values are coded assuming values determined as shown in FIG.

A procedure for decoding a bit stream created according to the third embodiment will be described with reference to FIG. FIG. 67 is a diagram illustrating a bit stream output method in the encoder according to the third embodiment.

First, in S40, the encoder performs visul
Outputs a bit stream according to the syntax of the Object sequence. The syntax of the Visual Object Sequence is the same as that shown in FIG. 14 of the first embodiment.

[0299] Next, proceeding to S41, the encoder performs visul Obje
Output a bitstream according to the syntax of ct. The syntax of the Visual Object is the same as that shown in FIGS. 15 and 16 of the first embodiment.

[0300] Next, proceeding to S42, the encoder performs the Video Obje
Output a bitstream according to the syntax of ct. The syntax of the Video Object is the same as that shown in FIG. 18 of the first embodiment.

Next, in S43, the head of the following bit stream is read, and the type of the following bit stream is determined. That is, the next bit stream is Vi
If it is deoObject_layer_start_code or short
In the case of t_video_start_markter, the process proceeds to S44, where decoding is performed as in the case of the first embodiment. (However, "fix
(it is determined that ed_intradc_q "does not exist and its value is 0.) Returning to S43, the flag" fixed_intradc_q "indicating the quantization method in the intra block indicates" fixed_intradc_q ".
When the value of “d_intradc_q” is 1, that is, in the first embodiment described above, when the quantization step of the DC coefficient in the intra block is determined by the value of intra_dc_precision,
Proceeding to S46, a bit stream starting from Video_Object_layer_start_code is output. Thereafter, the process proceeds to S47, where the output of the Video Object Plane is performed.

[0302] In this manner, the syntax for adaptively performing encoding is changed according to the value of "fixed_intradc_q". In the third embodiment, "fix
Although the method of changing the synth has been described using “ed_intradc_q”, a DC coefficient quantization method or a form of syntax for performing encoding may be selected using other means.

Next, a fourth embodiment of the present invention will be described. Here, the same encoding method / decoding method and encoder / decoder as those in the first, second, and third embodiments are provided, but the encoding method of the AC coefficient of the DCT coefficient is provided for them. Only differ. That is, in the first, second, and third embodiments, the DCT coefficient is
A plurality of sets of three parameters of N, LEVEL, and LAST are combined and encoded using VLC or the like. However, the corresponding VLC table corresponds to only some combinations and does not exist in the table. As for the combination, encoding is performed using the escape code as described above. However, if the bit precision per pixel is large, these VLC
In the encoding and the escape encoding, there is a possibility that sufficient efficiency may not be obtained. Therefore, in the fourth embodiment, a method of changing the encoding / decoding method according to the accuracy of the pixel of the input image. Will be described.

As described above, the quantized AC coefficient is encoded from (RUN, LEVEL, Last). Here, in the fourth embodiment, the obtained (RUN, Level, Last) is processed according to the following procedure. That is, the pixel precision of the image being encoded / decoded is other than 8 bits, and
Bit precision (not_8bit == 1 && bits
If _per_pixel> 8) Divide the value of Level by the determined number. That is, when the pixel precision of an image being encoded / decoded is n bits, division is performed by the number (n−8) +1. In other words, for 9 bits, division is performed by 2, 10 bits, 4 for 11 bits, and 16 for 12 bits.

The divided value is newly assigned to the Level, and the remainder generated when the Level is divided by (n−8) +1 is assigned to the DCT_resid.

After such processing is completed, each of the newly obtained (RUN, Level, Last) sets is coded by the above-described coding method, and the corresponding bit is output. Thereafter, the remainder obtained by the above-described processing is encoded with a fixed length of (n-8) bits.

[0307] A (RUN, Level, Last) decoding method according to this method will be described. That is, at the time of encoding the AC coefficient, the encoding method described above uses the VLC table or the escape code (RUN, Level).
l, Last). After that, when the image to be decoded has 8 bits or more, the following processing is performed for Level. That is, when the image to be decoded has n bits, the obtained Level is multiplied by (n-8) +1. That is, 2 for 9 bits, 4 times for 10 bits, 1
One bit is multiplied by 8 and 12 bits is multiplied by 16. Thereafter, the DCT_resid is read as fixed bits of (n-8) bits and added to the (n-8) +1 times Level. This value is replaced with the level read by the bit stream, and after inverse quantization, inverse DCT processing is performed.

Although this processing has been described only for the DC coefficient of the DCT, the processing can be applied to the case where the DC coefficient of the DCT is encoded by the same method as the AC coefficient.

FIG. 68 shows a configuration example of the syntax of the block layer to which this encoding method is applied. In FIG. 68, Fi
rst DCT coefficient is the difference L of the DCT DC coefficient.
evel, which occurs when the DC coefficient is coded in the same manner as the AC coefficient. If the pixel precision is 8 bits or more, the coded Level value is DCT_
Reconstructed using the value of resid.

As described above, DCT_resid is used to reconstruct Level when the pixel precision is 8 bits or more, and its value is a fixed value of (n-8) bits when the pixel precision of the input image is n bits. Encoded in length. Subsequent
The DCT coefficients are obtained by encoding the AC coefficients of the DCT by the method described above. When the pixel precision is n bits (8 bits or more), the code of (n-8) bits is further read and the DCT coefficients are read by the method described above. Decrypt the Level of

Although this processing has been described only for the AC coefficient of the DCT, the present invention can be applied to the case where the DC coefficient of the DCT is encoded by the same method as the AC coefficient.

As described above, DCT_resid is used to reconstruct the Level when the pixel precision is 8 bits or more, and its value is a fixed value of (n-8) bits when the pixel precision of the input image is n bits. Encoded in length.

The Subsequent DCT coefficients are DCT
Are encoded by the above-described method, and when the pixel precision is n bits (8 bits or more),
Read the (n-8) -bit code and use the DC
The level of the T coefficient is decoded.

Also, the fourth embodiment uses the DCT D
When the C coefficient is also reflected, the syntax is as shown in FIG. That is, the difference value of the DC coefficient DCT_DC_diffr
DCT_DC_siz representing the size of ential and DCT coefficients
e_luminance and DCT_DC_size_chrominance are encoded in the same manner as in 8-bit encoding even when the pixel precision is n bits (where n> 8), and (n-8) bits of information are added to correct Level It shall be.

[0315] By providing encoded data generated based on the encoding method according to the present invention through a recording medium or the Internet, image data having a gradation different from, for example, 4 to 12 bits can be provided to the user's desire. Can be provided accordingly.

Further, by recording a processing program based on the image encoding method on a recording medium, for example,
Image data having gradations of up to 12 bits can be easily generated by, for example, a personal computer.

[0317]

According to the present invention, the encoding device switches the required accuracy (number of bits) of the DC component coefficient of the orthogonal transform such as DCT in accordance with the required image quality. Can be changed, and the variable-length code table for DC component coefficients can be further extended, so that efficient coding can be performed.

On the decoding device side, the decoding method of the DC component coefficient is changed in accordance with the signal indicating the accuracy of the DC component coefficient of the orthogonal transform such as DCT to be transmitted, and the decoding method for the DC component coefficient is further changed. The variable length code table is extended, and the decoding of the coefficient becomes possible.

[Brief description of the drawings]

FIG. 1 Visu as a necessary term for understanding the present invention
al Object, video Object, Video Object Layer (VO
L), Group of Video Object Plane (GOV), Vid
eo Object Plane (VOP), Group of block (GO
FIG. 3B is a diagram for explaining macroblocks (MB) and blocks.

FIG. 2 is a diagram showing an arrangement of blocks in an MB in a 4: 2: 0 format.

FIG. 3 is a block diagram illustrating a configuration of an encoder that is an application example of the image encoding method and the image encoding device according to the first embodiment of the present invention.

FIG. 4 is a diagram showing a composite image including an image F1 as a background and an image F2 as a foreground.

FIG. 5 is a block diagram showing a detailed configuration of a VOP encoding unit inside the encoder shown in FIG. 3;

FIG. 6 is a block diagram showing a detailed configuration of a quantization circuit inside the VOP encoding unit shown in FIG. 5;

FIG. 7 is a flowchart for explaining an operation of a quantization step generator inside the quantization circuit shown in FIG. 6;

8 is a diagram showing a DC inside a VOP encoding unit shown in FIG.
FIG. 3 is a block diagram illustrating a detailed configuration of a T coefficient differentiator.

FIG. 9 is a block diagram showing a detailed configuration of an inverse quantizer inside the VOP encoding unit shown in FIG. 5;

FIG. 10 is a block diagram for explaining the operation of a DC predicted value selector inside a DCT coefficient differentiator having a detailed configuration shown in FIG. FIG. 4 is a diagram illustrating a positional relationship between blocks that refer to DC values.

FIG. 11 is a block diagram illustrating a configuration of a decoder that is an application example of the image decoding method and apparatus according to the first embodiment of the present invention.

FIG. 12 shows a VOP inside the decoder shown in FIG. 11;
It is a block diagram which shows the specific structural example of a decoding part.

FIG. 13 shows D in the VOP decoder shown in FIG. 12;
It is a block diagram which shows the detailed structure of a C coefficient inverse differencer.

FIG. 14 is a diagram illustrating the syntax of the bit stream output from the encoder shown in FIG.
Visual Object defined by sual Comitee Draft
It is a figure showing the syntax of Sequence.

FIG. 15 is a first division diagram showing the syntax of the Visual Object in a divided manner.

FIG. 16 is a second division diagram illustrating the syntax of the Visual Object in a divided manner.

FIG. 17 is a diagram illustrating the syntax of user_data.

FIG. 18 is a diagram illustrating the syntax of Video Object (VO).

FIG. 19 is a first division diagram showing the syntax of a video object layer (VOL) by dividing it.

FIG. 20 is a second division diagram illustrating the syntax of a video object layer (VOL) by dividing the syntax.

FIG. 21 is a third division diagram illustrating the syntax of a video object layer (VOL), which is divided.

FIG. 22 is a fourth division diagram illustrating the syntax of a video object layer (VOL) divided.

FIG. 23 is a diagram illustrating the syntax of a Group_of_Video Object Plane (GOV).

FIG. 24 is a first division diagram illustrating the syntax of a Video Object Plane (VOP) divided;

FIG. 25 is a second division diagram illustrating the syntax of a Video Object Plane (VOP), which is divided;

FIG. 26 is a third division diagram illustrating the syntax of a Video Object Plane (VOP) divided;

FIG. 27 is a fourth divided diagram illustrating the syntax of the Video Object Plane (VOP) divided.

FIG. 28 is a fifth division diagram illustrating the syntax of a Video Object Plane (VOP), which is divided;

FIG. 29 is a sixth division diagram illustrating the syntax of a Video Object Plane (VOP), which is divided;

FIG. 30 is a seventh division diagram illustrating a syntax of a Video Object Plane (VOP) divided;

FIG. 31 is an eighth division diagram illustrating the syntax of a Video Object Plane (VOP) divided;

FIG. 32 is a ninth division diagram illustrating the syntax of the Video Object Plane (VOP) divided;

Fig. 33 is a diagram illustrating the syntax of motion_shape_texture.

Fig. 34 is a diagram illustrating the syntax of combined_motion_shape_texture.

FIG. 35 is a first division diagram showing the syntax of a macro block divided.

FIG. 36 is a second division diagram illustrating the syntax of a macro block divided.

FIG. 37 is a third divided diagram illustrating the syntax of a macro block.

Fig. 38 is a diagram illustrating the syntax of mb_binary_shape_coding.

Fig. 39 is a diagram illustrating the syntax of motion_vector.

FIG. 40 is a diagram illustrating the syntax of a block.

Fig. 41 is a diagram illustrating the syntax of data_partitioning_motion_shape_texture.

FIG. 42 is a first division diagram illustrating the syntax of data_partitioning_I_VOP in a divided manner.

[Fig. 43] Fig. 43 is a second divided diagram illustrating the data_partitioning_I_VOP syntax in a divided manner.

FIG. 44 is a first division diagram illustrating data_partitioning_P_VOP syntax divided.

[Fig. 45] Fig. 45 is a second division diagram illustrating the syntax of data_partitioning_P_VOP by dividing the syntax.

[Fig. 46] Fig. 46 is a third divided diagram illustrating the data_partitioning_P_VOP syntax in a divided manner.

FIG. 47 is a flowchart for explaining the operation of a DC quantization step generator incorporated in an encoder as a specific example of the image signal encoding method and apparatus according to the second embodiment;

FIG. 48 shows a Video Object Pla used in the second embodiment.
It is the 1st division view which divides and shows the syntax of ne (VOP).

FIG. 49 shows a Video Object Pla used in the second embodiment.
It is the 2nd division figure which divides and shows the syntax of ne (VOP).

FIG. 50 is a diagram illustrating Video Object Pla used in the second embodiment.
It is the 3rd division view which divides and shows the syntax of ne (VOP).

FIG. 51 shows a Video Object Pla used in the second embodiment.
It is the 4th division view which divides and shows the syntax of ne (VOP).

FIG. 52 is a flowchart for describing a bit stream output method according to the third embodiment.

FIG. 53 is a diagram illustrating a configuration example of a slice in one frame.

FIG. 54 is a video object layer used in the third embodiment.
It is the 1st division view which divides and shows the syntax of er.

FIG. 55 is a video object layer used in the third embodiment.
It is the 2nd division figure which divides and shows the syntax of er.

FIG. 56 is a video object layer used in the third embodiment.
It is the 3rd division view which divides and shows the syntax of er.

FIG. 57 is a video object layer used in the third embodiment.
It is the 4th division view which divides and shows the syntax of er.

FIG. 58 is a video object layer used in the third embodiment.
It is the 5th division figure which divides and shows the syntax of er.

FIG. 59 is a video object layer used in the third embodiment.
It is the 6th division view which divides and shows the syntax of er.

[Fig. 60] Fig. 60 is a first division diagram illustrating the syntax of video_object_plane_highquality_header divided.

[Fig. 61] Fig. 61 is a second division diagram illustrating the syntax of video_object_plane_highquality_header, which is divided.

FIG. 62 is a third division diagram illustrating the syntax of video_object_plane_highquality_header, which is divided.

[Fig. 63] Fig. 63 is a first division diagram illustrating the syntax of the Video Object Plane, which is divided.

[Fig. 64] Fig. 64 is a second divided diagram illustrating the syntax of the Video Object Plane by dividing the syntax.

[Fig. 65] Fig. 65 is a third divided view illustrating the syntax of Video Object Plane.

FIG. 66 is a diagram showing values of some flags shown in the Video Object Plane.

FIG. 67 is a flowchart for describing a bit stream decoding procedure in the third embodiment.

FIG. 68 is a diagram illustrating the syntax of a block layer used in the fourth embodiment.

FIG. 69 is a diagram illustrating the syntax of a block layer when the fourth embodiment is also applied to the DC coefficient of the DCT.

FIG. 70 is a block diagram illustrating a configuration example of an encoder of MP @ ML in the MPEG system.

FIG. 71 is a diagram for describing properties of two-dimensional DCT.

FIG. 72 is a diagram for describing a method of converting a DC component of a DCT coefficient into a difference.

FIG. 73 is a block diagram showing a specific configuration of a DC coefficient differencer and a corresponding inverse differencer.

FIG. 74 is a diagram showing MP @ M in MPEG for decoding the encoded data output from the encoder shown in FIG. 70;
FIG. 3 is a block diagram illustrating a configuration example of an L decoder.

[Explanation of symbols]

3n VOP encoder, 135 quantizer, 307
DC coefficient quantizer, 308 quantization step generator

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Teruhiko Suzuki 6-35, Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Corporation F-term (reference) 5C057 AA06 CA01 CE10 EA02 EA07 ED07 ED09 EG08 EL01 EM04 EM09 EM13 EM16 GJ01 5C059 MA00 MA05 MA23 MA31 MB01 MB11 MB12 MB21 MC11 MC32 MC34 MC38 ME01 ME02 ME13 ME15 NN01 NN28 PP05 PP06 PP07 PP16 SS01 SS07 SS11 TA36 TA46 TB17 TB18 TC02 TC10 TD05 UA02 UA05 UA33 UA39 5J016 AA03 BA09 BA09 BC09

Claims (18)

[Claims] 1. An image coding apparatus for coding an image and outputting a resulting coded bit stream, comprising: an orthogonal transformation unit that performs a quadrature transformation according to a desired image quality of the image and a pixel precision indicating a pixel value per pixel Encoding accuracy determining means for determining the encoding accuracy of the DC component coefficient of the data; and quantizing means for quantizing the DC component coefficient of the orthogonal transform data according to the encoding accuracy determined by the encoding accuracy determining means. An image encoding device comprising: 2. The image encoding apparatus according to claim 1, wherein said encoding accuracy determination means determines the encoding accuracy of said DC component coefficient in video object layer units. 3. The encoding accuracy determining means according to claim 1, wherein said encoding accuracy of said DC component coefficient is determined from N to N + 3 bits when said image has N-bit accuracy. Image coding device. 4. The coding accuracy determination means according to claim 1, wherein said coding accuracy determination means determines the coding accuracy of said DC component coefficient from N to N + 2 bits when said image has N-bit accuracy. Image coding device. 5. The image encoding apparatus according to claim 1, wherein the DC component coefficient has a value for each of a luminance signal and a color difference signal. 6. An image coding method for coding an image and outputting a coded bit stream, comprising the steps of: performing orthogonal transform data conversion on the image in accordance with a desired image quality and a pixel accuracy indicating a pixel value per pixel; An image coding method comprising: determining the coding accuracy of a DC component coefficient; and quantizing the DC component coefficient of the orthogonal transform data according to the determined coding accuracy. 7. An image coding apparatus for coding an image and outputting a resulting coded bit stream, comprising: a main part of orthogonal transform data according to a pixel precision indicating a pixel value per pixel of the image; An image encoding apparatus, comprising: an encoding unit that separately encodes a residual part and a residual part. 8. An image coding method for coding an image and outputting a coded bit stream obtained as a result, comprising the steps of: performing orthogonal transform data conversion in accordance with pixel precision indicating a pixel value per pixel of the image; An image encoding method, wherein a main part and a residual part are separately encoded. 9. An intra-image and inter-image encoding process is performed,
And an image decoding device that orthogonally transforms the pixel data and decodes the image signal that has been subjected to the quantization process. An image decoding apparatus comprising: a decoding accuracy determining unit that adaptively determines a decoding accuracy used for the inverse quantization process of the inverse quantization unit according to a signal representing the accuracy and a pixel accuracy of the image. . 10. The image decoding apparatus according to claim 9, wherein said decoding accuracy determining means determines the decoding accuracy of said DC component coefficient for each video object layer. 11. The image decoding apparatus according to claim 9, wherein said decoding accuracy determining means determines the decoding accuracy of said DC component coefficient from N to N + 3 bits when said image has N-bit accuracy. apparatus. 12. The image decoding apparatus according to claim 9, wherein said decoding precision determining means determines the decoding precision of said DC component coefficient from N to N + 2 bits when said image has N-bit precision. apparatus. 13. The image decoding apparatus according to claim 9, wherein said DC component coefficient has a value for each of a luminance signal and a color difference signal. 14. An image decoding method for decoding an image signal which is subjected to intra- and inter-image encoding processing and pixel data subjected to orthogonal transformation and quantization processing, wherein: An image decoding method, wherein the decoding accuracy when the DC component coefficient is inversely quantized is adaptively determined according to the signal representing the accuracy of the DC component coefficient and the pixel accuracy of the image. 15. An image decoding apparatus for decoding an image signal subjected to intra-image and inter-image encoding processing, pixel data subjected to orthogonal transformation and quantization processing, and represents a pixel value per one pixel of the image. An image decoding apparatus comprising: decoding means for separately decoding a main part and a residual part of quantized orthogonally transformed data according to pixel precision used for the decoding. 16. An image decoding method for decoding an image signal which has been subjected to intra-image and inter-image encoding processing and pixel data subjected to orthogonal transformation and quantization processing, wherein a pixel value per one pixel of the image is obtained. An image decoding method characterized in that a main part and a residual part of quantized orthogonally-transformed data are separately decoded according to pixel precision used to represent the image decoding method. 17. An encoding accuracy of a DC component coefficient of orthogonal transformation data is determined according to a desired image quality of an image and a pixel accuracy indicating a pixel value per pixel, and the encoding accuracy is determined according to the determined encoding accuracy. An encoded data providing medium for providing an encoded bit stream obtained by quantizing a DC component coefficient of the orthogonal transform data. 18. An encoded bit stream obtained by separately encoding a main part and a residual part of orthogonal transform data according to pixel precision indicating a pixel value per pixel of an image. An encoded data providing medium characterized by the above-mentioned.