WO2006047936A1

WO2006047936A1 - A method for determining the condition in zero block prejudgment and for prejudging zero block

Info

Publication number: WO2006047936A1
Application number: PCT/CN2005/001787
Authority: WO
Inventors: Zhong Luo; Bin Song; Yilin Chang; Ningzhao Zhou
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2004-11-02
Filing date: 2005-10-28
Publication date: 2006-05-11
Also published as: CN1770863A

Abstract

This invention discloses a method for determining the condition in zero block prejudgment which comprises the steps of: obtaining the predictive difference image block between 4*4 block of current video image and the reference 4*4 block of preceding frame; calculating the sum of absolute difference corresponding to said predictive difference image block; according to the sum of absolute difference, calculating the quantization value of the DC coefficient which is integer transformed from the predictive difference image block by using integer transforming expression; obtaining the sufficient condition that the quantization value of said DC coefficient is equal to zero as the condition in zero block prejudgment. Accordingly, this invention also discloses a method for prejudging zero block. This invention can be applied to the processing of zero block prejudgment in H.264 compression encoding standard and can improve the compression encoding efficiency in H.264 standard.

Description

Method for determining zero block pre-judging condition and zero block pre-determination method

The present invention relates to video compression coding techniques in a multimedia communication system, and more particularly to a zero block pre-determination method in a compression coding process.

Background technique

The H.264 compression coding standard is the most advanced international standard for video compression coding. Its full name is the International Telecommunication Union - Telecommunication Standardization Organization H.264 standard (ITU-T H.264 Recommendation, International Telecommunication

Union-Telecommunication H.264 Recommendation), hereinafter referred to as the H.264 standard. The H.264 standard was developed by the ITU-T and the International Organization for Standardization (ISO) and the International Electrotechnical Commission's Moving Picture Experts Group (MPEG). The standard was developed in 1999 and is attended by equipment manufacturers, software developers, telecom operators, universities and scientific research institutions in many countries. It brings together the most advanced video compression coding technology in the world and finally became the ITU- in 2003. International standard for T.

Compared with the previous video compression coding standard technologies such as H.263+, H.263++ and MPEG-4 Simple Profile, H.264 has greatly improved performance and functions. Among them, many coding tool options have been added to the function, which makes it suitable for more application types and a wider range of applications. In terms of performance, at the same code rate (bit-rate), the image quality is peak signal-noise. Ratio (PSNR, Peak Signal-to-Noise Ratio, referred to as Peak Signal-to-Noise Ratio) As a reference, the H.264 standard doubles the previous H.263+/H.263++, MPEG-4 Simple Profile standard. , that is, the PSNR is increased by 3dB; that is, the network bandwidth required by the H.264 standard is higher than that required by the H.264+/H.263++, MPEG-4 Simple Profile standard when the same PSNR is obtained. Network bandwidth is reduced by 50%.

However, the H.264 standard has been significantly improved in terms of function and performance compared to the prior art standards, and the cost is high. The computational complexity of the H.264 standard for compression and decoding of video images is complicated. More computational complexity than standards such as H.263/H.263+ Adding several times, therefore, the premise of the large-scale popularization of the H.264 standard must be to develop a variety of processing methods with high processing efficiency to reduce the high computational intensity of the H.264 standard codec operation process. Calculate the computational complexity of the link so that the processing power of the multimedia processor required by the product can be reduced, thereby reducing the cost; or the multimedia processor can support higher encoding and decoding capabilities for a given processing capability, thereby improving product performance.

In the process of codec processing of the H.264 standard, there are multiple computational links with high computational complexity, such as Inter-frame prediction (Inter) and intra-frame prediction (Intra-frame prediction). Intra), Context-Adaptive Binary Arithmetic Coding. 1/4 pixel precision motion estimation (ME, Motion Estimation) and motion compensation (MC, Motion Compensation), multi-reference frame prediction (Multi- Hypothesis prediction ) and so on.

Especially in the intra prediction coding process and the interframe prediction coding process, the H.264 standard has a large processing strategy change with respect to the previous standards such as H.263/H.263+. Among them, in the process of encoding and decoding video images, the H.264 standard allows for prediction processing for each macro block (Mac block, hereinafter referred to as MB) in the current frame of the video image. The general prediction process first performs inter prediction, then performs intra prediction, and then compares the results of intra prediction and inter prediction, and selects the prediction mode with the highest coding efficiency, so that intra prediction and inter prediction are used. Predictive techniques are organically combined to achieve higher prediction efficiency and compression efficiency.

Among them, intra prediction is a unique processing attribute of the H.264 standard. For video image encoding process, intra prediction can maximize the spatial correlation of image data, thereby improving compression coding efficiency. In general video image compression coding processing, inter-prediction is more commonly used, and inter-frame prediction uses temporal correlation between video image sequences; however, many types of video images exist internally. Strong spatial correlation, so the H.264 standard designed the intra prediction method to make full use of the spatial correlation existing inside the video image to improve the compression coding efficiency.

Intra prediction in the H.264 standard refers to searching for a certain area (generally a rectangular block) currently to be encoded in a video image during encoding, and searching for the most similar and already encoded in the video image. Block, using the most similar block found in this search to predict the current encoding The block, the predicted result is the predicted residual. The similarity between different blocks is measured by SAD, Sum of Absolute Differences, that is, the block corresponding to the minimum SAD calculated by the block currently to be coded is taken as the closest block. The above image block may be a macroblock MB (ie, a block of 16×16 pixels size), or may be a smaller block than the macroblock MB, that is, the intra prediction in the H.264 standard is in MB or MB. On a smaller scale block.

In the video image compression coding process, whether it is inter-frame prediction or intra-frame prediction, it is based on MB or smaller-scale blocks (such as 16 x 8 blocks, 8 x 16 blocks, 8 x 8 blocks, 8 x 4 blocks, 4 X 8 blocks, 4 X 4 blocks, etc.) relative to the reference image block (ie, the image block used for prediction) to obtain a prediction residual (Presence Residue or Prediction Difference), the prediction residual result being an image of the same size Block; then transforming the prediction residual image block into a transform coefficient domain, and then quantizing the transform coefficients; and finally performing Zigzag scanning and entropy encoding on the quantized transform coefficients. It can be seen that if the optimization processing method is introduced in the process of transforming the prediction residual image block and quantizing the transformed transform coefficient, the image compression coding processing efficiency is greatly improved.

In the existing H.264 standard, the process of transforming the prediction residual image block and quantizing the transformed transform coefficient is performed. The optimization process is mainly considered from the following two aspects:

(1) Since the H.264 integer transform method is a transform method derived from the discrete cosine transform (DCT, Discrete Consine Transform) method, that is, the quotient of integers is used to approximate the DCT transform coefficients, thereby eliminating the floating point multiplication and division operations. By using the integer transform method instead of the DCT method, the transform efficiency can be greatly improved, and the transform process and the quantization process can be combined into one process. At the same time, since the image block size processed by the H.264 integer transform method is 4 x 4 blocks, if the H.264 integer transform is used and the prediction residual image block size is not 4 x 4 blocks, then the residual residual image needs to be taken. The block is divided into a plurality of 4 X 4 blocks, and then each 4 x 4 block is integer-transformed.

(2) Predetermining the zero block after the integer transform quantization (the so-called zero block is the abbreviation of the all-zero transform coefficient block, that is, the case where the transform coefficients of the 16 4 x 4 blocks in the MB are all zero), The prediction residual image block whose quantization result is zero has no influence on the image compression coding, so the transformation and quantization process for these zero blocks can be directly omitted, thereby improving the image pressure. The efficiency of the coding is reduced; and since the current block is the same position as the previous frame (for convenience of presentation, it is denoted as (0,0), the meaning of 0 is to indicate the relative motion displacement of the front frame block and the current block in the horizontal direction and the vertical direction. The prediction residuals are directly obtained between the corresponding blocks at positions 0, that is, the positions are exactly the same. Therefore, if the prediction residual image block transformation and the quantization result are all zeros, the subsequent motion search process can be avoided. Thereby, the motion search calculation amount of each image block can be saved, and the calculation amount of the entire inter prediction mode selection is saved correspondingly.

It can be seen that if it is possible to predetermine whether a quantized result of a prediction residual image block transformation is zero according to a certain condition with a certain probability (here, a certain small proportion of false positives is allowed), that is, the zero block after the integer transformation is determined in advance, It is possible to avoid a large number of zero-block transformation calculations and a partial calculation amount in the prediction mode selection process before the transformation. Therefore, in the H.264 standard, the pre-judgment processing of zero blocks is very important for improving the compression efficiency of video images.

The reason why the above transform transforms to generate a zero block is mainly because the quantization process essentially uses an exponential form factor of a quantization parameter (QP, Quantization Parameter) to divide the transform result, and then rounds the process, so as long as the dividend is sufficient (dividend, or Numerator is small enough, divisor, or denominator is large enough to get zero blocks. A large number of statistics on various video images show that in the image compression coding process, zero blocks exist in a large amount (of course, the number of zero blocks depends not only on the properties of the image itself, but also on the size of the QP, that is, the larger the QP is usually The more zero blocks are generated, so the zero block pre-determination technique has a considerable effect on the H.264 standard compression coding efficiency.

Specifically, there is currently no complete zero-block pre-determination technique that can be implemented in the H.264 standard, but since the H.264 standard is inherited from the H.263 standard, here 4 bar is in H. The zero-block pre-judging method applicable in the .263 standard is described as a cylinder.

In some implementations of the H.263 standard, a specific implementation of a zero block pre-determination method that can be used is as follows:

According to the mathematical properties of the DCT, each DCT coefficient is pre-determined by zero, as defined by the DCT: 0

, N-1 In the above formula, N denotes the image block size at which DCT is performed, that is, the image block processed by DCT is an N xN block; in the H.263 standard, since the image block processed by DCT is 8×8 blocks, N=8.

Where i = 0, l, 2, , N-1, j = 0, l, 2, , N-1; x(i, j) represents the number of the i-th row, the j-th column in the 8x8 block; i, j is a spatial domain indicator.

Where k = 0, 1 , 2, , Ν -1, 1 = 0, 1, 2, , N-1 in X(k, F}; X{k, l) represents the DCT coefficient; k, 1 is the transformation Domain indicator.

According to the above relationship, the following relationship can be introduced: 0,1,2,...; 7,/ = 0,1,2,·..,7 ( 2 )

The above formula uses the absolute value of ', which represents an upper bound estimate of the absolute value, so this relation can be used to give the zero pre-judgment condition of the 8 X 8 block DCT coefficient in the 263.263 standard:

Let Q denote the Quantization Level in the 263.263 standard, then 8 x 8

The quantized result of the DCT coefficient is: l ^U ) l ^{/ 22} integer quotient, ie if

Then, it can be determined that the quantized result of the DCT coefficient TM) is zero.

According to the above relationship (2), if there is:

(3) It can be determined that the quantized result of the DCT coefficient T t, /) is 0; and because k = 0, 1, 2, 7, 7, 0, 1, 2, 7, in the middle, so long as the above relationship (3) Satisfied

Λ-(0,0), (0,1), ..... ,χ(0,7)

Λ(1,0),Λ(1,1), .......,x(l,7)

χ = The DCT coefficients of the entire 8x8 block Λ(7,0), Λ-(7,1), . . . , χ(7,7) are quantized to be zero blocks in advance. Because the above prior art solutions are generally implemented only for the 263.263 standard, Combine 8 X 8 blocks, such as for each MB (16 X 16 blocks) in a P frame (predicted frame) or a B frame (backward predicted frame) in a video image sequence, after motion prediction, obtain prediction The residual image block is also 16 16 blocks, and then the block is divided into four 8×8 blocks, and then each block of 8 χ 8 blocks is subjected to zero block pre-decision processing of the above process, and the corresponding zero block can be obtained in advance.

Obviously, the zero-block pre-determination method described above in the Η.263 standard itself has the following drawbacks:

1) Since the Η.263 standard uses the DCT method, and the Η.264 standard uses the 4 4 integer conversion method; and the H.263 standard uses 8 x 8 blocks as the transformed unit image block, and the H.264 standard The 4 x 4 block is used as the transformed unit image block, so the zero block pre-determination method applicable in the H.263 standard is not well suited for the H.264 standard.

2) The zero block pre-determination method applicable in the H.263 standard also does not propose a practical and effective solution to avoid the motion search process in prediction mode selection. Because in the H.263 standard, there is no complex prediction mode selection in the H.264 standard, only inter prediction mode selection exists. Therefore, the zero block pre-judgment in the H.263 standard can save only the DCT of the image block. And the amount of quantization calculations behind, and the amount of computation for the motion search process in front of the DCT is small. Summary of the invention

The technical problem to be solved by the present invention is to propose a method for determining a zero block pre-determination component and a zero block pre-determination method, which is suitable for zero block pre-determination processing in the H.264 standard compression coding process, and improves the H.264 standard. Compression coding efficiency.

In order to achieve the above object, the present invention proposes a method for determining a zero block pre-judging condition, which is used for

The determination of the zero block pre-judgment condition in the H.264 compression coding process, including the steps:

A. Obtaining a prediction residual image block between the video image 4 x 4 block and the previous frame reference 4 x 4 block;

B. obtaining an absolute error sum corresponding to the prediction residual image block;

C. calculating, according to the absolute error sum, a quantized value of the DC coefficient value after the integer transform of the prediction residual image block by using an integer transform theoretical expression;

D. Finding a sufficient condition that the quantized value of the DC coefficient value is zero is taken as a zero block pre-decision strip

The displacement estimate

The value of each element value of the inter-predicted residual image block after the value search; in step C, the prediction residual image block is integer-transformed according to the following formula:

"1 1 1 1 χ (0, ϊ) χ (0, 2) χ (0, 3) "1 2 1 1

2 1 —1 -2 χ(1,2) χ(1,3) 1 1 -1 -2

1 -1 -1 1 (2,0) 2,1) Α2,2) χ(2,3) 1 -1 — 1 2

— 1 -2 2 -1, 3,0) χ(3,ϊ) χ(3,2) λ3,3) — 1 2 2 1 1 1— where Χ _4χ4 is the 4 x 4 block before the integer transformation, ^Χ 4" is an integer-transformed 4 x 4 block, Η is a 264.264 integer transformation matrix, which is a !! transposed matrix;

The DC coefficient value after integer conversion of the prediction residual image block in step C

∑x( ,n); the quantized value of the DC coefficient value is

/«=0

Where (0,0) represents the quantized value of the DC coefficient value, Q _M ≡QP"«? 6, Q _{E ≡} QP/6, QP

— 1, x<0

The quantity parameter and QP e [0,51]; function sign(x) = 0, x = 0 f value is selected by the video encoder;

1, x>0 The sufficient condition for making the quantized value of the DC coefficient value zero in step D is:

The 3⁄4 AG) value is determined by the following process:

A(Q , i,j) = M{Q _M , r), where i and j represent the rows and columns of the elements in the 4 x 4 block, respectively

0 for ( ) = {(0,0),(0,2),(2,0),(2,2)} coordinates, where 0≤z', _ ≤3; r is a function of i, j: r = 1 for ( ) = {(U), (1, 3), (3, 1), (3, 3)};

2 other 13107,5243,8066

11916, 4660, 7490

10082,4194,6554

M is a matrix: M =

9362,3647,5825

8192,3355,5243

7282, 2893, 4559

Since 4, 0, 0) ( ) is (o, ₀₎ and _r =0 is obtained, the Q _M row in the M matrix is _searched , and the 0th column element is used as the value of β _Μ , ο, ο).

Correspondingly, the present invention also proposes a zero block pre-determination method for the zero block pre-determination process in the 264.264 compression coding process, including the steps:

(1) Determine the absolute error corresponding to the prediction residual image block and SAD < (1 - /) · 2 ¹⁷⁺ / 2 _Μ , 0, 0) are zero block pre-judgment conditions; where Q _M≡ QP, " 6, Q _E≡ QP/6, QP is the quantization parameter and QPe [0,51;], f value is selected by the video encoder;

(2) for the current 4x4 block in the current image frame, find the reference 4x4 block of the initial position of the previous frame, and obtain the absolute error sum corresponding to the prediction residual image block between the current block and the initial position reference block;

(3) determining whether the absolute error and the zero block pre-judging condition are satisfied, and if so, recording the position information of the current 4x4 block in the current image frame in the zero block list; otherwise, going to (4);

(4) For the current 4x4 block, complete the motion search in the previous frame, find the reference 4x4 block of the optimal position of the previous frame, and obtain the absolute error corresponding to the prediction residual image block between the current block and the optimal position reference block. with;

(5) In the process of performing integer transform and quantization on the prediction residual image block between each 4x4 block in the current image frame and the corresponding reference 4x4 block in the previous frame, respectively, determining whether the position information of each 4×4 block is recorded or not In the zero block list, for the 4×4 block whose position information is recorded in the zero block list, the prediction residual image block between the reference block and the corresponding reference block of the previous frame is not subjected to integer transform and quantization processing; The 4x4 block recorded in the zero block list is subjected to integer transform and quantization processing on the prediction residual image block between the corresponding reference block and the previous frame.

The step (4) and (5) further includes the steps of: determining the absolute error in the step (4) and whether the zero block pre-judging condition is satisfied, and if yes, applying the current block to the current image frame The location information is recorded in the zero block list; otherwise it is not recorded. The location information refers to row and column coordinate information of the current block in the current image frame. Correspondingly, another zero block pre-determination method proposed by the present invention is used for zero block pre-determination processing in the H.264 compression coding process, including the steps:

(S1) determining the absolute error corresponding to the prediction residual image block and SAD < (1 - /) · 2 ¹⁷⁺ IA(Q _M , 0, 0) is a zero block pre-judgment condition; wherein Q _{M ≡} QP mod 6, Q _{E ≡} QP/6, QP is the quantization parameter and QPs [0, 51], f value is selected by the video encoder;

(S2) dividing the current block of the current image frame into the 4×4 blocks in the current inter prediction mode;

(53) For each 4x4 block divided by the current block, respectively find the reference 4 X 4 block of the previous frame initial position, and respectively obtain the prediction residual image between each 4x4 block and the corresponding initial position reference 4x4 block. The absolute error sum of the block;

(54) for the absolute error and the 4x4 block satisfying the zero block pre-judgment condition, the position information in the current image frame is recorded in the zero block list, and the zero block is not satisfied for the absolute error Predicting the conditional 4x4 block execution step (S5);

(S5) performing a motion search of the previous frame for each 4×4 block that does not satisfy the zero block pre-judging condition, and finds the reference 4x4 block of the previous frame optimal position, respectively, The absolute error sum corresponding to the prediction residual image block between each 4x4 block and the corresponding optimal position reference 4x4 block;

(S6) determining, in the process of performing integer transform and quantization on the prediction residual image block between each current block and the previous reference frame in the current image frame, determining whether the location information of each 4x4 block divided by the current block is Both are recorded in the zero block list, and if so, the prediction residual image block between each 4x4 block divided by the current block and the corresponding reference 4x4 block of the previous frame is not subjected to integer transform and quantization; otherwise, the current block is Each of the divided 4x4 blocks is subjected to integer transform and quantization processing with respect to the prediction residual image block between the corresponding reference 4×4 blocks of the previous frame.

The steps (S5) and (S6) further include the steps: for the absolute error in the step (S5) and the 4x4 block satisfying the zero block pre-judging condition, the position information in the current image frame is recorded in the In the zero block list, 4x4 blocks that do not satisfy the zero block pre-judgment condition are not recorded.

The inter prediction mode described in the step (S2) includes Interl6x l6, Interl6 x 8, Inter8 x 16, Inter 8 x 8, Inter 4 x 8 and Inter 8 x 4; The step (S5) and (S6) further includes the steps of: determining whether the inter prediction mode is Inter 8 8, Inter 4 χ 8 and Inter 8 x 4, ^ is, go to step 3⁄4 (S6); No-j'J-->J-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

The invention can achieve the following beneficial effects:

1) According to the mathematical theory of 4x4 integer transform in the H.264 standard, the derivation of the determined zero block pre-judging condition can be well applied to the zero block pre-determination in the 4x4 integer transform process in the H.264 standard. The judgment condition has good discriminant accuracy, which can better guarantee the image quality of H.264 video images.

2) The zero block pre-determination method proposed by the present invention according to the determined zero block pre-judging condition can be applied to any inter-prediction mode in the H.264 standard, and only the unit image block in the non-Inter 4x4 prediction mode is divided into 4x4. The block is processed, and the reference 4x4 block reference 4 X 4 block of the previous frame initial position is found for each 4x4 block at present, and the absolute error and SAD of the prediction residual image block between the current block and the initial block are calculated, and the SAD is satisfied. In the case of a conditional condition, the motion search of the current 4x4 block is stopped; if the SAD does not satisfy the pre-judgment condition, the motion search of the current 4x4 block is continued until the reference block of the optimal position is found; for the SAD, the pre-judgment condition is satisfied. 4x4 block, the position information in the current frame is recorded in the zero block list, and the 4x4 block recorded in the zero block list is not subjected to integer transform for the prediction residual image block between the reference block and the corresponding reference block of the previous frame. And quantization processing (for all 4x4 block position information divided by unit image blocks in non-Inter 4x4 prediction mode, it is recorded in the zero block list. The prediction residual image block between each 4x4 block and the corresponding reference block of the previous frame is not subjected to integer transform and quantization processing, which can save inter prediction mode selection and integer transform and quantization in the H.264 standard. Calculate the amount, which can improve the efficiency of compression coding in the H.264 standard. DRAWINGS

1 is a flow chart showing the main implementation principle of the method for determining the zero block pre-judging condition of the present invention; FIG. 2 is a flow chart showing the principle of the zero block pre-determination method of the present invention in the Inter 4x4 inter-frame prediction mode;

3 is a flow chart showing the implementation principle of the zero block pre-determination method of the present invention in Interl6 χ 16, Inter 16 x 8, Inter8 x 16, Inter 8 x 8, Inter 4 x 8 and Inter 8 x 4 inter prediction modes. detailed description

First, the method for determining the zero block pre-judging condition of the present invention proposes a zero block pre-judging condition that can be applied to the H.264 standard 4 x 4 integer transform. The proposed zero block pre-judging condition has the following requirements:

Simple and practical, the judgment accuracy is high, and the zero block false positive rate is low; and there are two cases for false positives: one is to take the image block which should be zero block, and it is not judged to be zero block (ie, missed judgment). This kind of misjudgment makes the degree of improvement of coding efficiency affected. The other is to misjudge the image block which is not zero block, and the misjudgment will lead to a significant drop in image quality. Therefore, the proposed zero block pre-requirement should be required. Judging conditions can minimize the occurrence of these two misjudgments.

The specific implementation of the method for determining the pre-judging condition of the zero block of the present invention will be described in detail below with reference to the accompanying drawings.

Because in block-based video image coding standards such as H.263, image conversion is usually performed using near-optimal KL (Kahunen-Loeve, a statistical transformation), for a random vector (the image block is a random vector), Through the KL transform, "whitening" can be performed ("whitening" is to make its autocorrelation matrix into a diagonal matrix); for any image block, the KL transform can be used to obtain the best approximate representation. However, in practical applications, because the K-L transform is more complicated, the discrete cosine transform (DCT) is usually used instead of the K-L transform, and it is proved that under certain conditions (general images are satisfied), DCT is the optimal approximation of the K-L transform. However, the DCT method is used for image transformation, and some of the transformation matrix elements are irrational. If floating point operation is used, the error between the transformation and the inverse transformation will be caused. Therefore, the H.263 standard requires DCT transform and inverse transform to have sufficient accuracy to ensure the recovery accuracy of the decoded image, and the computational complexity of both DCT transform and inverse transform is also large. Thus, the H.264 standard proposes to transform an image block using a 4 X 4 integer transform.

Among them, in the H.264 standard, the definition of 4 x 4 integer transform is as follows:

1 1 1 1 x(0,0) x(0,l) χ(0,2) χ(0,3)

2 1 -1 -2 | xiiO) χ(1,ϊ) χ(1,2) χ(1 )

^4x4 =Η·Χ _4χ4 ·Η ^Τ

1 -1 -1 1 χ(2,0) χ(2,1) λ 2,2) χ(2,3) 1 1-1 1

1 -2 2 -1 3⁄40) χ(3,ΐ) χ(3,2) χ(3,3) 1 1-2 1 In the above formula (the coefficient block before transformation) and ^Χ (the coefficient block after transformation) respectively represent the coefficient matrix before and after the inter prediction residual value transformation, Η is the integer transformation matrix in the 264.264 standard, ⁷ is 11 Transpose matrix. It can be seen from the above formula that there is no multiplication in the 4x4 integer transform operation in the 264.264 standard, only the addition and shift operations avoid the precision matching problem of floating-point operations in the forward and reverse transforms, and reduce the operation of the decoder. the complexity. In addition, the 4x4 integer transform in the 264.264 standard facilitates the optimization of parallel processing of algorithms using Single Instruction Multiple Data (SIMD).

At the same time, in order to avoid division in the quantization process, the H.264 standard quantizes the transform coefficients after integer transformation using the following quantization formula:

X _q (i, j) = sign {X _T (i, j)} |x _r (, j]A{Q _M , i, j) + 2 ¹⁷⁺ &)» (l7 + Q _E )] Q _{E ≡} QP/6, Qp is the quantization parameter and QPe[ ,51J , letter

number

Where i and j represent the row and column coordinates of an element in a 4x4 block, respectively, where 0 ≤ _ζ ·, _/· ≤ 3; X _T ( ) and X^'j) respectively represent the integer transform coefficients before and after quantization Value; aW, > is

The quantized variation factor given by the H.264 standard, whose value depends on the spatial position of the transform coefficient; and the value of the parameter f is selected by the video encoder, and the typical value of f is [0, 0.5], which The effect is to maintain the quantization accuracy in the truncation process (the truncation process is what is commonly referred to as rounding).

Where ^ = ^M Q _M ' (6) where r is a function of i, j in the above equation (6), defined as follows:

0 for ( ) = {(0,0),(0,2),(2,0),(2,2)}

r = 1 for ( ) = {(1,1), (1,3), (3,1), (3,3)}

, 2 other ( ₇ )

M is a 6x3 matrix, defined as follows: 13107,5243,8066

11916, 4660, 7490

10082,4194,6554

M =

9362,3647,5825

8192,3355,5243

7282, 2893, 4559

(8)

Therefore, the quantization transformation factor is calculated from the value of (ij)

The process is: firstly r is calculated according to equation (7), and in accordance with formula (6) the query expression (8), to give Q values of the _M rows of the matrix M, the first r columns of elements as the quantized transform factor).

The standard conversion process and the quantization process of the 4x4 integer transform in the H.264 standard have been described above. Please refer to FIG. 1, which is a flow chart of the main implementation principle of the method for determining the condition of zero block pre-judgment of the present invention; the main implementation process is as follows:

Step S10, obtaining a prediction residual image block between the video image 4x4 block and the previous frame reference 4x4 block;

Step S20, obtaining an absolute error corresponding to the prediction residual image block and SAD, SAD = where (" θ represents each of the inter prediction residual image blocks after the displacement estimation search

Step S30, according to the SAD corresponding to the prediction residual image block obtained in step S20, using the foregoing integer transformation theoretical expression, obtaining the DC coefficient value after the integer transformation of the prediction residual image block obtained in step S10 Continue to find the DC coefficient

Value; the quantized value of ^(0,0) is:

Step S40, according to the integer transform property in the H.264 standard, the absolute value of the DC coefficient value _r (0, 0) is the largest among the 4×4 integer transform coefficients, and it is assumed here that if the transform coefficients are quantized If the DC coefficient value (0,0) of the value ^X _T ( ) is 0, then all the AC coefficient values in ^Χ τ( ) are also zero, that is, the transform coefficient value of the entire image block is all zero, and the block is zero. Piece. Therefore, according to this principle, as long as the quantized value of the DC coefficient value in the transform coefficient of one prediction residual image block is zero, the quantization result of the entire image block can be satisfied to be zero. It can be seen from the above formula (5) that the sufficient condition that the quantized value of the DC coefficient value ^(0, 0) is zero is:

And because ΣΣ ^Λ '("ζ,"< ∑∑ \x{m, n , the amount of DC coefficient value Χ _τ (0,0) can be obtained.

//=0 w/=0 Η=0 /"=0 The sufficient condition that the value is zero is: ( , "〗 <(l- /)'2 ¹⁷⁺ /A{Q _M ,0,0) (10 The amount to the right of the inequality in inequality (10) is actually a threshold, which can be represented by a symbol TZB, and ZB represents a Zero Block.

T _ZB = (lf) - 2 ^m ^ / A (Q _M , 0, 0)

The above inequality is replaced by amplification, and the condition is strengthened. That is, if the equation (9) is a sufficient condition that the quantized value of the DC coefficient value (0, 0) is zero, the equation (10) is also a sufficient condition.

According to the above description principle, the decision condition of the above formula (10) can be used as the zero block pre-judgment condition of the integer transform in the H.264 standard. Using the zero block pre-judgment condition jx{m,n]<{lf)-2 ^m& /A{Q _m ,0,6) determined for the integer transform in the H.264 standard determined above.

«=o-=, the present invention further proposes a corresponding zero block pre-determination method, which is directed to the complex inter-prediction mode selection process in the 264.264 standard, using the above-determined zero block pre-judging condition, The block decision processing is advanced (that is, not only after the inter-frame prediction is processed, but also before the transform processing, the zero-block decision is made, but the zero-block decision process is advanced to the inter-prediction mode selection process), so that Can save inter prediction mode The calculation amount of the selection process further improves the compression coding efficiency of the H.264 standard.

The specific implementation process of the zero block pre-determination method of the present invention will be described in detail below with reference to the accompanying drawings.

Through the derivation of the above process, the zero block pre-judgment condition in the H.264 standard is determined as

The left item of the formula is 4 X

4 Predict the absolute error and SAD value of the residual image block. Therefore, for the Inter 4 x 4 inter prediction mode in the H.264 standard, only the SAD of the 4 x 4 prediction residual image block needs to be calculated, and the zero block pre-determination can be performed. In Inter 4 x 4 motion prediction, the current 4 x 4 block of the current frame first performs a motion search starting from the 4 x 4 reference block of the corresponding position of the previous frame, which is called the initial position. If at the initial position, the SAD of the prediction residual image block between the current 4×4 block and the 4×4 reference block of the previous frame initial position has met the above-described zero block pre-judgment condition, then even if the reference block of the initial position is not motion prediction The optimal position does not matter, because the reference block relative to the optimal position of the previous frame has a smaller SAD than the SAD of the reference block relative to the initial position, so if the SAD of the reference block relative to the initial position can satisfy zero The block pre-judgment condition, the SAD of the reference block of the optimal position must also satisfy the zero block pre-judgment condition, and thus the prediction residual image block between the reference blocks of the current 4×4 block relative to the optimal position of the previous frame can be determined. After quantization, it is an all-zero block, and the prediction residual image block between the reference blocks relative to the initial position of the previous frame is quantized to be all zero blocks. Since the initial position and the optimal position are all zero blocks, there is no effect on the compression coding result, so the 4 x 4 reference block of the initial position of the previous frame can be used directly as the basis of the zero block prediction, without necessarily The reference block to find the optimal position of the previous frame. Therefore, as long as the SAD calculated with respect to the reference block of the initial position of the previous frame satisfies the zero block pre-judging condition, the operation of the current 4 X 4 block can be ended. The search is performed, and it is not necessary to search for the reference block of the optimal position of the previous frame, thereby completing the motion search process of the H.264 standard.

Please refer to FIG. 2, which is a flow chart of the implementation principle of the zero block pre-determination method in the inter 4x4 inter-frame prediction mode of the present invention; the main implementation process is as follows:

Step S100, determining an absolute error corresponding to the prediction residual image block and the SAD according to the above process

Block pre-judgment condition; where Q _M ≡QP"w^6, Q _E ≡QP/6, QP is the quantization parameter and QP e [0,51], f value is selected by the video encoder, and other parameters are determined by reference to the above description;

Step S110, in the Inter4 4 inter prediction mode, find a reference 4x4 block of the initial position of the previous frame for the current 4x4 block in the current image frame;

Step S120, determining SAD ₀ , ₀ corresponding to the prediction residual image block between the current block and the initial position reference block;

In step S130, the SAD obtained in step S120 is determined. ,. Whether the zero block pre-judgment condition SAD < T _ZB is satisfied, if yes, step S140 is performed; otherwise, step S150 is performed;

Step S140, terminating the motion search process of the current 4x4 block, and recording the position information of the current 4x4 block in the current image frame in the zero block list ¹ ^, where x and y represent the row of the 4×4 block position in the current frame, Column coordinates, then step S180;

Step S150, for the current 4x4 block, continue to complete the motion search process of the previous frame, and find the reference 4×4 block of the optimal position of the previous frame;

Step S160, obtaining an SAD corresponding to the prediction residual image block between the current 4x4 block and the optimal position reference block;

Step S170, it is determined whether the SAD obtained in step S160 satisfies the zero block pre-judgment condition SAD<T _ZB , and if yes, go to step S140; otherwise, the recording is not performed, and step S180 is continued;

Step S180, determining, in each of the 4x4 blocks, performing integer transform and quantization on the prediction residual image block between each 4x4 block in the current image frame and the corresponding reference block in the previous frame. The position information is recorded in the block list ^Lx .y zero, the processing procedure for the information recorded in the zero position of the block in the list ^L w 4 X 4 block, the prediction residual between the current frame its corresponding reference block The image block is not subjected to integer transform and quantization processing, and the all-zero value is directly taken to the 4×4 block; for the 4×4 block whose position information is not recorded in the zero block list ¹ ^, between the reference block corresponding to the previous frame The prediction residual image block is subjected to integer transform and quantization processing.

The above-described zero block pre-determination method of the present invention can greatly reduce the number of calculations of the SAD in the 4×4 block motion search for the zero block pre-decision algorithm in the Inter 4 χ 4 inter prediction mode in H.264, and also reduces the integer transform. And the amount of quantization, and has no effect on the coding performance of the Inter 4 χ 4 inter prediction mode.

Since the Η.264 standard supports coded blocks of different sizes and shapes, the optional inter prediction mode includes the following inter prediction modes in addition to the Inter 4 x 4 inter prediction mode: Interl6 χ

16, Interl6 x 8, Inter8 x 16, Inter 8 ^ 8, Inter 4 x 8 and Inter 8 x 4. However, the prediction process of these inter prediction modes and the prediction process of the Inter 4 x 4 inter prediction mode are similar, and there is a process in which the previous frame is searched from an initial position, so if the zero block advance can be performed at the initial position In the judgment, it is possible to decide whether to terminate the search process for the current block according to the judgment result. Therefore, the block larger than 4×4 in other inter prediction modes can be divided into a plurality of 4×4 blocks by means of the zero block pre-determination method in the Inter 4×4 inter prediction mode described above, and then For each 4 x 4 block, apply the above zero block pre-judgment condition SAD =

To make a judgment. Please refer to FIG. 3, which is an implementation of the zero block pre-determination method of the present invention in Interl6 x16, Inter 16 x 8, Inter8 χ 16, Inter 8 χ 8, Inter 4 χ 8 and Inter 8 x 4 inter prediction modes. Principle flow chart; The main implementation process is as follows:

〉:〉1·^(" ” Step S200, determining the absolute error corresponding to the prediction residual image block and SAD = "=."'=.' < (1- /)·2 ^17+3⁄4 /A(Q _M , 0,0) = 2 ^ is the zero block pre-judgment condition in the _{H 264} integer transform process; where Q _M ≡QP"«^6, Q _E ≡QP/6 ₅ Qp is a quantization parameter and QPe [0, 51], f value is selected by the video encoder, and other parameters are determined by referring to the above description; Step S210, in the above non-Inter4x4 inter prediction mode, the current image is The current block of the frame in various inter prediction modes is uniformly divided into a plurality of 4x4 blocks, so that each of the 4x4 blocks in different inter prediction modes can be pre-zeroed using the above-described zero block pre-judging condition; Step S220, For each 4x4 block divided by the current block, respectively find the reference 4x4 block at the initial position of the previous frame;

Step S230, respectively obtaining SAD^ corresponding to the prediction residual image block between each 4x4 block and the corresponding initial position reference 4x4 block;

Step S240, determining whether the SAD _G , _Q obtained in step S230 satisfies the above-mentioned zero block pre-judgment condition SAD < ^T ZB , performing step S250 for the 4x4 block satisfying the condition; performing step S260 for the 4x4 block not satisfying the condition;

Step S250, for the 4x4 block that satisfies the condition, record the position information in the current image frame in the zero block list ¹ ^, where x and y represent the row and column coordinates of the 4x4 block position in the current frame, and then perform the steps. S290;

Step S260, for each 4 χ 4 block of SAD _{0j0 that} does not satisfy the above-mentioned zero block pre-judgment condition SAD < ^, respectively complete the motion search process of its previous frame to find the reference 4 X 4 block of its previous frame optimal position. ;

Step S270, respectively obtaining the SAD corresponding to the prediction residual image block between each 4x4 block and the corresponding optimal position reference 4x4 block;

Step S280, it is judged whether the SAD obtained in step S270 satisfies the zero block pre-judgment condition SAD < ^T ZB , and for the 4x4 block satisfying the zero block pre-judgment condition, the process goes to step S250; for the case that the zero block pre-judgment condition is not satisfied 4x4 block is not recorded, directly executing step S290;

Step S290, in performing integer transform and quantization on the prediction residual image block between each current block and the corresponding reference block in the previous frame in the current image frame, determining each of the current block partitions.

Whether the location information of the X 4 block is recorded in the zero block list ¹ ^, and if so, step S300 is performed; Otherwise, step S310 is performed;

Step S300, terminating the motion search process for the current block, and performing the integer transform and quantization processing on the prediction residual image block between each 4×4 block and the corresponding reference block of the previous frame divided by the current block, directly taking Is an all-zero value; for example: In Inter 16 x 8 inter-frame prediction mode, if 8 4 4 4 blocks divided by each current 16 X 8 block in this mode satisfy the above-mentioned zero block pre-judgment condition, The current 16 x 8 block motion search process is terminated prematurely, that is, the 16 x 8 block can be directly judged to be an all zero block.

Step S310: Perform integer transform and quantization processing on the prediction residual image block between each 4×4 block that is divided by the current block and the corresponding reference block of the previous frame.

Here, in order to further ensure the encoding quality of the H.264 image, it is also determined between steps S280 and S290 whether the current inter prediction mode is Inter 8 x 8 , Inter 4 x 8 and Inter 8 x 4, if Step S290 is performed; otherwise, the prediction residual image block between each 4×4 block divided by the current block and the corresponding reference block of the previous frame is separately subjected to integer transform and quantization processing.

In order to verify the technical effect of the inventive scheme, three typical standard video image sequences "foreman" (sports intensity), "news" (medium exercise), "claire" (small degree of motion) are used. The frame is tested. The image format is uniformly QCIF (Quarter Common Interchange Format), Y:U:V is 4:2:0; the experimental platform is Pentium IV, the computer with the main frequency is 2GHz, and the value of parameter f is chosen to be 0.79. .

Table 1 below shows the experimental results of the zero block pre-determination method when using different quantization parameters QP in the Inter 4 χ 4 inter prediction mode. It can be seen from the table that with the increase of QP, after using the method of the present invention, all zero 4 x 4 blocks can account for more than 80%, of course, the proportion of all zero blocks is still the content of the image sequence and the QP. Value related. In addition, after adopting the scheme of the present invention, the zero block false positive rate is very low, so that the coding performance of the .264 standard is basically unaffected. Table 1: Percentage of all zero blocks for different image sequences in Inter 4 x 4 inter prediction mode

QP Claire News Foreman 18 46.41% 23.42% 2.1%

28 72.22% 48.87% 21.14%

38 87.35% 68.29% 41.76% As can be seen from Table 1, Figure ^: The more intense the degree of motion, the smaller the proportion of all zero blocks (because the all-zero residual image block of inter prediction represents the change between the corresponding blocks of the preceding and succeeding frames. Small, this means that the image is not moving much). In the video image, except for the action movie video, the amount of motion of other video images is medium or even low, but even for action movie video, the lens with a really large amount of motion is only a part of it, like a general multimedia video. Images, such as video conferences, are video images with a low amount of motion. That is to say, for a general video application, there are many all-zero blocks in the video image sequence.

Table 2 below gives experimental results of applying the zero block pre-determination method of the present invention to all inter prediction modes of the H.264 standard, which compares the use of the zero block pre-determination method and the zero block pre-determination method of the present invention. Coding performance, compression efficiency and coding speed. Table 2: Improvement of coding speed after using the method of the invention Image sequence QP PSNR (dB) Number of coding bits (bits) Coding speed (frames/second) Coding speed is adopted before adoption, before adoption, after adoption, before use Percentage foreman 28 35.58 35.51 1604880 1567872 36.8 38.5 4.6% foreman 38 29.01 28.96 422728 431176 39.9 44.6 11.8% news 28 36.59 36.53 835312 805648 45.2 55.4 22.6% news 38 29.14 29.12 267944 263688 48.1 66.7 38.7% claire 28 39.85 39.78 409008 396424 48.3 71.2 47.4 % claire 38 33.07 32.98 202688 203544 52.6 83.3 58.4% As can be seen from Table 2, after using the method of the invention for all inter prediction modes, the coding speed can be increased by up to 58.4% compared with before the method of the invention is used; The average PSNR of the decoded restored image is less than O.ldB, that is, from the preservation of the image quality, there is almost no decrease, and the number of encoded bits is slightly reduced (i.e., the compression coding efficiency is slightly improved). It can be seen that, after adopting the method of the invention, the coding speed of the H.264 standard will be greatly improved during the implementation process, which can reduce the cost of the multimedia communication product and improve the performance of the multimedia communication product, thereby improving The market competitiveness of multimedia communication products.

The above description is only a preferred embodiment of the present invention, and it should be noted that those skilled in the art can also make several improvements and retouchings without departing from the principles of the present invention. It should be considered as the scope of protection of the present invention.

Claims

Rights request

A method for determining a zero block pre-judgment condition for determining a zero block pre-judging condition in an H.264 compression coding process, comprising:

A. Obtaining a prediction residual image block between the 4x4 block of the video image and the 4x4 block of the previous frame reference;

D. A sufficient condition for zeroing the quantized value of the DC coefficient value is obtained as a zero block pre-judgment condition.

2. The method for determining a zero block pre-judgment condition according to claim 1, wherein the absolute error sum in step B is (m, «, ,; c(m, ") represents a displacement estimate. Searching for each of the inter-frame prediction residual image blocks °3⁄4°;

In step C, the prediction residual image block is integer-transformed according to the following formula:

χ(0,2) χ(0,3) "1 2 1 1 χ(12) χ(1,3) 1 1 — 1 22 χ(2,2) Χ2,3) 1 — 1 -1 2

(3,2) ^(3,3) - 1 - 2 1 -1 - where Χ _4χ4 is the 4x4 block before the integer transform, ^χ 4χ4 is the 4x4 block after the integer transform, Η is the 264.264 integer transform matrix, #是11的转置矩阵; Step c

The quantized value of the DC coefficient value is:

Where: v(0,0) represents the amount of DC coefficient value ^^Q?mod6 , Q _E ≡QP/6, QP is the quantization parameter and QP€ [0,51]; Function ^^ =| 0, Λ- = 0, · f value is selected by the video encoder; the sufficient condition for making the amount of the DC coefficient value ^1⁄4 in step D is: 1⁄2n,n) < (1 - /).2 ¹⁷ /A(Q _M ,0,0).

3. The method for determining a zero block pre-judgment condition according to claim 2, wherein the A(Q _M , G, G) value is determined by the following process:

A(Q _m , i,j) = M(Q , r), where i and j represent the rows and columns of the elements in the 4x4 block, respectively

0 for ( ') = {(0,0),(0,2),(2,0),(2,2)}, where 0≤z'J≤3; r is a function of i, j: V-1 for ( ')-{(U), (1,3), (3,1), (3,3)}

2 other

13107,5243,8066

11916, 4660, 7490

10082,4194,6554

M is a matrix: M =

9362,3647,5825

8192,3355,5243

7282, 2893, 4559

Query the first matrix Q M _M line, the first element as 0 (g _M, o, o) values.

4. A zero block pre-determination method for zero block pre-determination processing in the H.264 compression coding process, characterized in that it comprises the steps of:

(1) Determine the absolute error corresponding to the prediction residual image block and the SAD < (ΐ- )· 2 ¹⁷⁺ 3⁄4 / A(Q _M , 0,0) is a zero block pre-judgment condition; where Q _{M ≡} QPm _0i /6 , Q _{E ≡} QP/6 , QP is the quantization parameter and QPe[0, 5l], f value is selected by the video encoder;

(4) For the current 4x4 block, complete the motion search in the previous frame, find the reference 4×4 block of the optimal position of the previous frame, and obtain the corresponding prediction residual image block between the current block and the optimal position reference block. Absolute error sum;

(5) Pre-reference between the 4x4 block and the previous frame in the current image frame. In the process of integer transform and quantization, the residual image block is respectively determined whether the position information of each 4x4 block is recorded in the zero block list, and the position information is recorded in the 4x4 block in the zero block list, and the previous frame is The prediction residual image block between the corresponding reference blocks is not subjected to integer transform and quantization processing; for the 4x4 block whose position information is not recorded in the zero block list, the prediction residual image block between the reference block and the corresponding reference block of the previous frame is performed. Integer transform and quantization processing.

The zero block pre-determination method according to claim 4, wherein the step (4) and (5) further comprise the step of: determining whether the absolute error in the step (4) and the Zero block pre-judgment condition, if yes, the position information of the current block in the current image frame is recorded in the zero block list; otherwise, it is not recorded.

The zero block pre-determination method according to claim 4 or 5, characterized in that the step

(I) The establishment process of the zero-block pre-judgment condition includes the following steps:

(II) obtaining a prediction residual image block between the video image 4 X 4 block and the previous frame reference 4x4 block; (12) obtaining an absolute error sum corresponding to the prediction residual image block;

(13) determining, according to the absolute error sum, a quantized value of the DC coefficient value after the integer transform of the prediction residual image block by using an integer transform theoretical expression;

(14) A sufficient condition for making the quantized value of the DC coefficient value zero is obtained as a zero block pre-judgment condition.

The zero block pre-determination method according to claim 4 or 5, wherein the position information refers to row and column coordinate information of the current block in the current image frame.

8. A zero block pre-determination method for zero block pre-determination processing in the H.264 compression coding process, characterized in that the method comprises the steps of:

(S1) determining the absolute error corresponding to the prediction residual image block and SAD < (1-/).2 ¹⁷⁺ /4δ _Μ , 0, 0) are zero block pre-judging conditions; wherein Q _{M ≡} QP"iOi/6, Q _E ≡QP/6, QP is the quantization parameter and QPe [0,51], f value is selected by the video encoder;

(S2) dividing the current block of the current image frame into various 4x4 blocks in the various inter prediction modes;

(S3) For each 4x4 block divided by the current block, respectively find the reference 4 X 4 block of the previous frame initial position, and respectively obtain the prediction residual image between each 4x4 block and the corresponding initial position reference 4x4 block. The absolute error sum of the block;

(S4) for the absolute error and the 4 X 4 block satisfying the zero block pre-judgment condition, The position information in the pre-image frame is recorded in the zero block list, and the step (S5) is performed for the absolute error and the 4x4 block that does not satisfy the zero block pre-judgment condition;

(S6) determining, in the integer transform and quantization process, the prediction residual image block between each current block and the previous reference frame in the current image frame, determining the position of each 4×4 block divided by the current block Whether information is recorded in the zero block list, and if so, the prediction residual image block between each 4×4 block divided by the current block and the corresponding reference 4x4 block of the previous frame is not subjected to integer transform and quantization; otherwise An integer transform and quantization process is performed on each of the 4x4 blocks divided by the current block and the prediction residual image block between the corresponding reference 4x4 blocks of the previous frame.

9. The zero block pre-determination method according to claim 8, wherein the step (S5) and (S6) further comprises the steps of: for the absolute error in the step (S5) and satisfying the zero block pre- The conditional 4x4 block is recorded in the zero block list for its position information in the current image frame, and is not recorded for the absolute error and the 4x4 block that does not satisfy the zero block pre-judgment condition.

The zero block pre-determination method according to claim 8 or 9, wherein the establishing process of the zero block pre-judging condition in the step (S1) comprises:

(S11) obtaining a prediction residual image block between the video image 4 X 4 block and the previous frame reference 4x4 block; (S12) obtaining an absolute error sum corresponding to the prediction residual image block;

(513) calculating, according to the absolute error sum, a quantized value of the DC coefficient value after the integer transform of the prediction residual image block by using an integer transform theoretical expression;

(514) A sufficient condition for zeroing the quantized value of the DC coefficient value is obtained as a zero block pre-judgment condition.

The zero block pre-determination method according to claim 8 or 9, wherein the position information refers to row and column coordinate information of the 4 X 4 block in the current image frame.

The zero block pre-determination method according to claim 8 or 9, wherein the inter prediction mode in the step (S2) comprises Interl6x l6, Interl6x8, Inter8 x 16, Inter 8 ^x 8, Inter 4 x 8 and Inter 8 x 4; The step (S5) and (S6) further includes the steps of: determining whether the inter prediction mode is Inter 8x 8, Inter 4 x 8 and Inter 8 x 4, if yes, proceeding to step (S6); The prediction residual image block between each 4x4 block divided by the current block and the corresponding reference block of the previous frame is subjected to integer transform and quantization processing.