JP2001119304A - Image coding device - Google Patents

Abstract

(57) [Problem] To reduce a difference in image quality between intra-frame coding and inter-frame coding. SOLUTION: An input image 101 is a video interface 1
02 is input to the difference image generation unit 1 as the original image data 103.
04, which is processed by the DCT unit 105, the quantization unit 106, and the encoding unit 108, and converts the quantization error data 124 in the quantization unit 106 into an error calculation unit.
118, and an error calculation unit 11 in a quantization step setting unit 107.
Quantization step 12 so that the error value 132 of 8 is almost constant
Configure to give 1.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding device and an image encoding method for encoding moving image data.

[0002]

2. Description of the Related Art MPEG2 (I
SO / IEC 13818-2). In MPEG2, images are classified into two types, intra-frame and inter-frame, and a high compression rate is realized by using motion compensation in inter-frame images. Since the generated code amount is small in the inter-frame image, the image quality of the intra-frame coding and the inter-frame coding may fluctuate.

[0003] If the difference between the fluctuations is large, the image looks like undulating, and the image quality is degraded. As a solution to this problem, Japanese Patent Application Laid-Open No. Hei 10-32816 has been proposed, and this phenomenon is described under the name of GOP pulsing.

[0004]

SUMMARY OF THE INVENTION Japanese Patent Application Laid-Open No. 10-3281
Japanese Patent Application Laid-Open No. 6-29564 is based on the premise that it is performed in two paths, and changes the period (GOP structure) of the change between intra-frame coding and inter-frame coding in order to avoid GOP pulsing.

However, in the conventional method as described above, real-time realization is difficult, and when the GOP structure is changed, the period of the intra-frame coded image fluctuates. Or, it becomes difficult to handle in editing.

Accordingly, an object of the present invention is to provide a new means for calculating a measure of image quality degradation in an encoded image, and to control intra-frame encoding and inter-frame encoding by controlling a quantization step using the measure. An object of the present invention is to provide an image encoding device that reduces the difference in image quality between encoded images and reduces GOP pulsing.

[0007]

In order to achieve the above object, the present invention provides an error calculating means for calculating a conversion error in a converting means for performing an irreversible conversion, and an error calculating means for converting a irreversible part by the converting error. The conversion parameters can be adjusted.

With this configuration, it is possible to detect image quality deterioration, and as a result, GOP pulsing can be reduced by parameter control using image quality deterioration.

Further, the conversion error is obtained by a quantization error in a quantization means for irreversible conversion.

With this configuration, it is possible to detect image quality degradation with a relatively simple configuration, and GOP pulsing can be reduced as a result of parameter control using image quality degradation.

In addition, the conversion error is obtained by using a difference between the conversion error and a result obtained by inverting the conversion result.

With this configuration, it is possible to detect image quality deterioration in a form close to an actual image, and as a result, GOP pulsing can be reduced by parameter control using image quality deterioration.

Further, the quantization step is controlled so that the obtained conversion error becomes a constant value. Since the conversion error becomes constant, the image quality within a frame and between frames becomes substantially uniform, and GOP pulsing can be reduced.

Further, a target conversion error is set before the intra-frame coding, and the target conversion error is updated by an average of the conversion errors of the frame after the intra-frame coding.

According to this configuration, the bit generation amount can be controlled by the target conversion error before the intra-frame encoding, and furthermore, the intra-frame encoding image and the inter-frame encoding to be encoded thereafter are controlled. The image quality can be made constant, and GOP pulsing can be reduced.

[0016]

Embodiments of the present invention will be described below with reference to the drawings.

(Embodiment 1) FIG. 1 shows a first embodiment of an image coding apparatus according to the present invention. FIG.
This is an example applied to a G encoder.

In the MPEG encoder shown in FIG. 1, an input image 101 is divided into processing units called macroblocks of 16 × 16 via an image input means video interface 102. The processed original image data 103 is sent to the motion detection unit 113 and the difference image generation unit 104.

The motion detecting section 113 searches the reference frame memory 115 for an image closest to the original image data 103, and locates the position in the form of a motion vector 112 in the predicted image generating section 1.
Send to 14.

The predicted image generation unit 114 generates a predicted image 120 based on the data in the reference frame memory 115 using the given motion vector 112.

The difference image generation unit 104 generates an image of the difference between the original image data 103 and the prediction image 120. At this time, when the difference data is large or when the image of intra-frame encoding (I pictur
In the case of e), since the motion compensation is not performed, the original image data is output to the DCT unit 105 as it is, and otherwise, the difference data is output to the DCT unit 105.

At this time, information as to whether or not the motion compensation has been performed is sent as a mode signal 130 to the encoding unit 108 and the image reconstruction unit 111.

The DCT unit 105 performs a discrete cosine transform on the data, and supplies the result to the quantization unit 106. The quantization unit 106 calculates D
The data subjected to the CT processing is quantized. Since the quantization step is basically an integer division, a remainder generally occurs. The remainder is provided to the error calculator 118. Also,
The quantization step is set by the quantization step setting unit 107.

The encoding unit 108 encodes the output of the quantization unit 106 and outputs it as an output bit stream 117 to the outside.

The inverse quantization unit 109 performs inverse quantization on the result of the quantization in a given quantization step 121. This inverse quantization step is a multiplication, and the result of the inverse quantization does not match the input of the quantization unit 106.

The inverse DCT processing section 110 performs an inverse quantization process. Here, the result of performing the inverse DCT processing is approximately DCT
Although it matches the input of the unit 105, it does not completely match the original data due to the quantization of the quantization unit 106 and the like.

The image reconstructing unit 111 generates a reconstructed image 116 from the result of the inverse DCT processing unit 110, the predicted image 120 and the mode 130. This corresponds to the inverse transformation of the difference image generating unit 104, and the reconstructed image 116 substantially matches the original image data 103.

The reconstructed image 116 is stored in the reference frame memory 115.
And is used as a reference image for the subsequent image frames.

The error calculation section 118 calculates an error value 132 using the quantization error data 124 from the quantization section 106.

The error value 132 can be calculated by, for example, the maximum quantization error data or the weighted average of the quantization error data.

The quantization step setting section 107 calculates the quantization step 121, and the calculation method will be described later.

In the MPEG encoder shown in FIG.
The configuration is almost the same as that of a normal MPEG encoder except for the error calculator 118.

The MPEG encoding is usually performed using motion compensation using the blocks shown in FIG. 1. However, the encoding is performed only within a frame for random access and error resistance, that is, without performing motion compensation. Contains the frame to be converted.

As described above, for an image encoded only within a frame, the difference image generation unit 104 gives the original image to the DCT unit 105 as it is, and the image reconstruction unit 111 also performs the inverse DCT operation.
The output of the processing unit 110 is used as it is.

Also, MPEG is irreversible compression, especially since data information is dropped in the quantization unit 106.
The encoding unit 108 is changed by changing the quantization step.
Can control the amount of bits generated.

When the quantization step is increased, the encoding unit
The amount of bits generated at 108 is reduced, and as a result, the data amount of the output bit stream 117 can be reduced.

In an MPEG encoder, a target bit rate is usually given in advance, and it is only necessary to keep the target bit rate and encode with as high an image quality as possible.

As described above, the process of changing the quantization step to maintain the bit rate is called rate control, and there are various methods for achieving high image quality. It is an object of the present invention to reduce the difference in image quality between intra-frame and intra-frame coding in a still image.

Therefore, the present embodiment provides a method for controlling the quantization step so as to equalize the difference from the original image (mainly caused by a quantization error).
This is to reduce an unnatural change in image quality in P. For this purpose, a quantization step is calculated based on the difference from the original image. These processes are performed by the quantization step setting unit 107.

FIG. 2 is a flowchart showing the operation of the quantization step setting unit 107.

In STEP 201 in FIG. 2, two works Q0 and R0 are initialized as initial settings.

In step 202, Q0 is set as a quantization step 121.

In STEP 203, the error data from the error calculation section 118 is fetched. Let this data be R.

In STEP 204, Q0 is changed based on the difference between R0 and R.

More specifically, if R is larger than R0, Q0 is made smaller, and conversely, if R is smaller than R0, Q0 is made larger. Where Q
Since 0 is used as a quantization step in SETP202, Q0
Is large, the image quality tends to deteriorate, that is, R tends to increase, and conversely, if Q0 is small, the image quality tends to improve, that is, R tends to decrease.
That is, it is controlled so as to approach.

At STEP 205, the bit generation amount 131 is encoded by the encoding unit 108.
Take in from.

In STEP 206, R0 is changed based on the difference between the bit generation amount 131 fetched in STEP 205 and the target.

The reason is that if the bit generation amount 131 is larger than the target, R0 is increased (error is increased).
Thus, if the bit generation amount is smaller than the target, control is performed so as to reduce R0.

By changing R0 in this manner, when the bit generation amount is larger than the target, the quantization step is increased as a result, whereby the bit generation amount can be reduced.

Returning from STEP 206 to STEP 202, the above processing is repeated for each macroblock.

As described above, in this embodiment, since the quantization step is determined so that the quantization error is constant, the difference in image quality between frames can be made uniform, and as a result, As a result, a more natural encoded image can be obtained.

Although the control using only the quantization error and the bit generation amount is shown here, the quantization step to be set is modulated by using a statistic such as the variance of the pixels of the macro block. Such a method may be used. Further, these processes are performed in units of macroblocks, but may be performed in larger units.

(Embodiment 2) FIG. 3 is a block diagram of an encoder according to Embodiment 2 of the present invention. The encoder of FIG. 3 is almost the same as that of FIG. 1, but differs from the first embodiment only in the operation of quantization step setting section 307.

FIG. 4 shows the operation of the quantization step setting unit 307. Although the basic concept is the same in the second embodiment, it differs from the first embodiment in that different operations are performed according to intra-frame coding and inter-frame coding.

The SETP 401 in FIG. 4 is an initial setting, and initializes R0 and Q0.

At STEP 402, it is determined whether the frame to be encoded is an intra-frame or an inter-frame.

If intra-frame encoding is performed,
403 is executed, and R0 is changed according to the bit generation amount encoded so far. At this time, if the bit generation amount is large, R0 increases, and if the bit generation amount is small, R0 decreases.

In STEP 404, Q0 is set as a quantization step.

In STEP 405, an error is obtained from the error calculation unit 118. This is R.

In STEP 406, Q0 is changed according to the difference between R and R0.

This is an operation similar to that of STEP 204 of the first embodiment, and has an effect of keeping R, which is an error, constant.

At STEP 407, it is determined whether or not the end of the frame is reached. If not, the process returns to STEP 404. Thus, unlike Embodiment 1, R0 is not updated.

In STEP 407, if the end of the frame is reached, the process proceeds to STEP 408.

In STEP 408, it is determined whether or not the frame is the last of the intra-frame coding.

If not intra-frame coded, STEP 40
Return to 2.

If the image is an intra-frame coded image, STEP
Move to 409.

In STEP 409, R0 is reset based on the value of R of the frame, and the process returns to STEP402.

As described above, in the second embodiment, the first embodiment
In contrast to this, R0 is updated at the end of the intra-frame coded image at the average R of the encoded image, and immediately before the intra-frame encoding, R0 is updated from the bit generation amount.

Therefore, when performing inter-frame encoding, an error of approximately the same level as that of the previously encoded intra-frame encoded image, that is, a substantially uniform image is obtained. At the beginning, a target error, that is, an image quality level is selected according to the bit generation amount.

Therefore, if the control as in the second embodiment is performed, the picture quality of the intra-frame coding and the inter-frame coding can be made substantially uniform, and the GOP pulsing can be reduced while controlling the bit generation amount.

(Embodiment 3) FIG. 5 is a block diagram of an MPEG encoder according to Embodiment 3 of the present invention.

In the third embodiment, most of the blocks are the same as in the first embodiment, but the quantization unit 506, the error calculation unit 518, and the quantization step setting unit 507 are different. The other blocks are exactly the same as those having the same numbers in FIG.

The quantization section 506 differs from the quantization section 106 of FIG. 1 in that it does not output quantization error data, but is otherwise the same.

The error calculator 518 is significantly different from the error calculator 118 in FIG. 1 and calculates an error value 532 using the original image data 103 and the reconstructed image 116 as inputs.

At this time, the error value 532 is calculated by, for example, the sum of the absolute values of the differences between the original image data 103 and the reconstructed image 116, or the sum of the squares of the differences.

The quantization step setting section 507 operates according to the flow shown in FIG. 3 or FIG. At this time, they are exactly the same except that an error value 532 is used instead of the error value 132.

The major difference as described above is that the error calculation unit
There is only 518, and the error value 532 more directly indicates the difference between the original image and the reconstructed image. Since the reconstructed image is the same as the image decoded by the decoder in MPEG, an index indicating the image quality degradation can be obtained more directly. Instead of the error due to the quantization error data in the first and second embodiments, Can be used.

Therefore, as in the first and second embodiments, the image quality of the intra-frame encoding and the inter-frame encoding is substantially uniform, and there is an effect of reducing GOP pulsing.

[0079]

As described above, according to the present invention, since the quantization step is controlled based on the error in the irreversible transform unit (DCT processing unit or quantization processing unit), the image quality of each frame can be made uniform. GOP pulsing can be reduced, so that the apparent image quality is improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of an MPEG encoder according to a first embodiment of the present invention.

FIG. 2 is a flowchart of an operation of a quantization step determining unit according to the embodiment;

FIG. 3 is a block diagram of an MPEG encoder according to a second embodiment of the present invention.

FIG. 4 is a flowchart of an operation of a quantization step determining unit according to the embodiment;

FIG. 5 is a block diagram of an MPEG encoder according to a third embodiment of the present invention.

[Explanation of symbols]

 101 Input image 102 Video interface 105 DCT unit 106 Quantization unit 107 Quantization step setting unit 118 Error calculation unit 307 Quantization step setting unit 518 Error calculation unit

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C059 KK01 KK47 MA00 MA04 MA05 MA23 MC11 MC31 NN01 PP01 PP04 SS21 TA46 TB04 TC36 TD12 UA02 5J064 AA01 BA13 BA15 BB01 BB03 BC01 BC02 BC16 BD01

Claims (8)

[Claims] An image input means for dividing moving image data into a plurality of processing units, and a conversion means for performing irreversible conversion based on parameters given to the processing units divided by the image input means. Encoding means for encoding the result of the conversion means, conversion error calculation means for calculating the conversion error of the irreversible conversion by the conversion means, and the conversion error calculation means based on the result of the conversion error calculation means. An image coding apparatus comprising: a parameter setting unit that changes a parameter. 2. The apparatus according to claim 1, wherein said transforming means includes DCT means for performing discrete cosine transform, and quantizing means for performing quantization in a given quantization step, and said parameter setting means performs a quantization step for said quantizing means. 2. The image encoding device according to claim 1, wherein the image encoding device is provided. 3. The conversion means comprises: a motion detection mode setting means for determining whether or not to perform motion detection for each frame; and a reconstruction mode setting means for determining whether to perform image reconstruction for each frame. Frame storage means; motion detection means for searching the reference frame storage means for data that best matches the processing unit input by the image input means; and generating a predicted image from the result of the motion detection means. A prediction image generation unit, a difference image generation unit that calculates a difference between the input data and the prediction image, a mode selection unit that selects one of the difference image and the input image, DCT means for performing a discrete cosine transform on data, and performing quantization based on a given quantization step for a result of the DCT means Decoding means, inverse transform means for performing inverse quantization and inverse discrete cosine when the image reconstruction is designated by the reconstruction mode setting means, and data and mode of the inverse transform means 2. The image processing apparatus according to claim 1, wherein said selecting means comprises a sum with said predicted image or an image reconstructing means for directly writing in said reference frame, wherein said parameter setting means gives a quantization step to said quantizing means. Image encoding device. 4. The image coding apparatus according to claim 2, wherein said error calculating means calculates said conversion error based on an error caused by quantization of each data of said quantizing means. 5. The image encoding apparatus according to claim 3, wherein said error calculating means calculates said transform error based on a difference between a result of said inverse transform means and an input to said DCT means. apparatus. 6. A target error is set for said quantization means, and said quantization means gives a quantization step so that said conversion error obtained by said conversion error calculation means becomes a target value. The image encoding device according to claim 4 or 5, wherein 7. The quantizing means sets a target conversion error based on the code amount generated by the encoding means at the time of intra-frame encoding, and after intra-frame encoding, 7. The image encoding apparatus according to claim 6, wherein the target conversion error is reset based on the value of the conversion error obtained by the conversion error calculation unit during the intra-frame encoding. 8. A step of dividing moving image data into a plurality of processing units, a step of performing irreversible conversion based on parameters given to the divided processing units, and encoding a result of the conversion. An image coding method, comprising: converting a conversion error of the irreversible conversion; and varying a parameter of the conversion based on a calculation result of the conversion error.