CN101867818B - Selection method and device of macroblock mode - Google Patents

Abstract

The invention discloses a selection method and a device of a macroblock mode. The method comprises the steps of comparing the coding costs of all coding modes of a current macroblock, determining the optimal macroblock coding mode, and adopting the pixel precision stratified selection method for part of inter-frame modes when calculating the coding costs; and a novel comparison algorithm can also be adopted when comparing the special modes-skip mode and other coding modes of the macroblock. The method can significantly reduce the computational complexity and greatly improve the computational parallelism while keeping the bit rate and the coding quality unchanged basically, thereby being applicable to real-time software and hardware video coding.

Description

Method and device for selecting macroblock mode

technical field

The present invention relates to the technical field of video coding, in particular, to a fast selection method and device for macroblock mode selection of video compression coding. the

Background technique

In each video coding standard, such as H.264/AVC (Advanced Video Coding), AVS (AudioVideo coding Standard Advanced Audio Video Coding Standard) all stipulate the code stream structure and decoding process, but do not describe the specific coding method. Mode selection is the core technology in video coding, so it is a hotspot of academic research. the

Since most coding standards use a macroblock as the basic unit of coding, the mode selection is based on the coding mode of the macroblock. According to the existing technology, the mode selection is to compare the coding cost functions among the various modes, and select the mode with the smallest coding cost function as the final mode of the current macroblock. The cost function of encoding is the rate-distortion cost (RDCost: Rate-Distortion cost) of the corresponding encoding mode when rate-distortion optimization (RDO: Rate-Distortion-Optimization) is adopted. Since the calculation of the rate-distortion cost is very complicated, some people have proposed other The method of calculating the cost function, but the amount of calculation is still very large, especially when the motion estimation accuracy is high, the complexity of doing motion estimation with high precision at one time is very high. Moreover, for special modes, such as skipping (SKIP) mode, the fast algorithm for mode selection does not use the cost function for comparison calculation, which will greatly reduce the accuracy of mode selection. the

Contents of the invention

The object of the present invention is to provide a method and device for fast selection of a macroblock mode in view of the deficiencies in the prior art. the

The purpose of the present invention is achieved by the following technical solutions:

A method for selecting a macroblock mode, first comparing the encoding costs of the first type of inter-frame mode under low-precision motion estimation, and selecting the mode with the best encoding cost as the selection result of the first type of inter-frame mode, and then selecting High-precision motion estimation is performed in the mode to obtain the optimal coding cost of this mode under high-precision motion estimation. The encoding cost is then compared with the encoding costs of the optimal intra mode and the optimal second-type inter mode to select a mode with a better encoding cost. The first type of inter-frame mode here refers to the inter-frame coding mode that requires motion search, the second type of inter-frame mode refers to the inter-frame coding mode that does not require motion search, and the better coding mode refers to the encoding The coding mode selected by the small cost or other judging criteria combined with the coding cost. the

An apparatus for the selection method of the macroblock mode. It includes: a low-precision pixel first-type inter-frame mode selector, a high-precision pixel motion estimator, an intra-frame mode selector, a second-type inter-frame mode selector and a mode decision device. Among them, the output terminal of the low-precision pixel first-type inter-frame mode selector is connected to the input terminal of the high-precision pixel motion estimator, and outputs the optimal coding mode in the first-type inter-frame mode and the coding cost of the macroblock in this coding mode . The output terminal of the high-precision pixel motion estimator is connected to the mode decision device, which outputs the corresponding mode and the coding cost of the macroblock in the coding mode after the high-precision motion estimation. The output terminal of the intra-frame mode selector is connected to the mode decision device, which outputs the optimal intra-frame mode and the coding cost of the macroblock in this coding mode. The output terminal of the second-type inter-frame mode selector is connected to the mode decision device, which outputs the optimal coding mode and the coding cost of the macroblock in the second-type inter-frame mode. One input terminal of the mode decision device is connected to the high-precision pixel motion estimator, one input terminal is connected to the intra-frame mode selector, and the other input terminal is connected to the second-type inter-frame mode selector, and the best of the three input modes is output model. the

A method for selecting a macroblock mode. The SKIP mode of a macroblock is selected by combining the threshold judgment of the absolute value of the predicted residual block with the comparison of the encoding cost function, that is, when the absolute value of the predicted residual block in the SKIP mode is less than a threshold determined by the size of the macroblock quantization parameter, and the encoding cost of SKIP is better than the encoding cost of the optimal mode in other modes except SKIP mode, the current macroblock selects SKIP mode; otherwise, the current macroblock selects SKIP mode The optimal encoding mode among other modes except mode. Here, a better coding mode refers to a coding mode selected with a smaller coding cost or other criteria combined with the size of the coding cost. the

A device used in the selection method of the above-mentioned macroblock mode, which includes a residual absolute value comparator, a first encoding cost calculator, a second encoding cost calculator, and a mode selector. Wherein, the output terminal of the residual absolute value comparator is connected to the mode selector, and outputs a signal indicating whether the residual absolute value is greater than a threshold. The output terminal of the first encoding cost calculator is connected to the mode selector, and outputs the optimal mode among other modes except SKIP mode and the encoding cost of the macroblock in this mode. The output terminal of the second coding cost calculator is connected to the mode selector, and outputs the coding cost of the SKIP mode. The three input ends of the mode selector are respectively connected to the residual absolute value comparator, the first encoding cost calculator and the second encoding cost calculator. Output the final selected encoding mode. the

The beneficial effect of the present invention is that the present invention can still accurately select the optimal macroblock coding mode under the condition of greatly reducing the computational complexity, and at the same time, the fine judgment of the SKIP mode improves the accuracy of the selection of the SKIP mode sex. the

Description of drawings

Fig. 1 is the structural block diagram of the selection device of the macroblock mode of embodiment 4 of the present invention;

Fig. 2 is the structural block diagram of the selection device of the macroblock mode of embodiment 5 of the present invention;

Fig. 3 is a structural block diagram of an apparatus for selecting a macroblock mode according to Embodiment 6 of the present invention. the

Detailed ways

The purpose and effect of the present invention will become more obvious according to the present invention in detail according to the accompanying drawings and embodiments. the

The technical proposal of the present invention mainly selects the coding mode with the smallest coding cost among various modes according to the coding cost function to improve the coding performance of the video coding system. the

Below is the concrete implementation method of the present invention:

Example 1:

Taking the AVS standard as an example, in the AVS standard, macroblock modes are divided into two categories: inter-frame coding mode and intra-frame coding mode. The inter-frame coding mode can be divided into 16×16, 16×8, 8×16, and 8×8 modes according to the size of the sub-block. In addition, there are special coding modes between frames, such as the P slice. SKIP mode, SKIP and DIRECT mode in B. The intra-frame coding mode is predictively coded according to the size of 8×8 blocks, and there are 5 kinds of intra-frame coding modes: horizontal mode, vertical mode, lower-right mode, lower-left mode, and DC mode. For each macroblock, it is necessary to select a mode with high coding efficiency among these coding modes for coding. The method for selecting the macroblock mode according to claim 1 is described below by taking the selection of the macroblock mode in the B slice in the AVS standard as an example. the

The specific implementation steps are as follows:

Step 1: Use the following cost function to select a mode with the smallest coding cost among the inter-frame modes 16×16, 16×8, 8×16, and 8×8. The encoding cost of the selected macroblock mode is recorded as minCost_inter_int. the

Cost_inter_int＝SAD_int+λ×(mode_bits+motion_bits)

Among them, SAD_int is the absolute value sum of the residual block after integer pixel precision motion search in the current mode, mode_bits is the number of bits required for the mode information of this mode to be transmitted in the code stream, and motion_bits is the bits required for the motion vector to be transmitted in the code stream λ is 0.67×QP, where QP refers to the quantization parameter of the current macroblock. the

Step 2, calculate the encoding cost of the mode selected in step 1 after 1/4 pixel precision motion search

minCost_inter_qua＝min_SAD_qua+λ×(mode_bits+motion_bits)

Where min_SAD_qua is the absolute value sum of the residual block after the 1/4 pixel motion search of the mode selected in step 1, mode_bits is the number of bits required for the mode to be transmitted in the code stream, and motion_bits is the motion vector transmitted in the code stream The number of bits required, λ is 0.67×QP, where QP refers to the quantization parameter of the current macroblock. the

Step 3, using the following cost function to select an intra-frame mode with the smallest encoding cost among the five intra-frame modes. The encoding cost of the selected macroblock mode is denoted as minCost_intra. the

Cost_intra＝SAD_intra+λ×mode_bits+H

Wherein SAD_intra is the absolute value sum of prediction residual blocks after intra prediction in the current mode, and mode_bits is the number of bits required for the mode information of this mode to be transmitted in the code stream. λ takes 0.67×QP. H is the correction amount of the intra-frame cost, which is taken as 18 here. the

Step 4, compare the encoding cost of the mode selected in step 1 and step 3, that is, the size of minCost_inter_qua and minCost_intra, if minCost_inter_qua is larger, set the mode selected in step 2 as a candidate mode, otherwise set the mode in step 1 Set to candidate mode. the

Step 5: Compare the coding cost of the candidate mode selected in step 4 with the coding cost of the optimal coding mode in the inter-frame special mode, where the special mode refers to DIRECT mode and SKIP mode. Select the mode with the least coding cost as the final coding mode of the current macroblock. The encoding cost of the special mode is calculated as follows. the

Cost_special＝SAD_special+λ×mode_bits

Among them, SAD_special is the absolute value sum of the residual block predicted by the special mode, mode_bits is the number of bits required for the mode information of this mode to be transmitted in the code stream, and λ is 0.67×QP, where QP refers to the quantization parameter of the current macroblock . the

Example 2:

The method for selecting the macroblock mode in claim 6 is described below by taking the method for selecting the macroblock mode in the P slice in the AVS standard as an example. the

The specific implementation steps are as follows:

Step 1: Use the following cost function to select a mode with the smallest encoding cost from other candidate modes except SKIP mode. The coding cost of the selected macroblock mode is recorded as minCost_normal. the

Cost_normal＝η×num_nozero+λ×(mode_bits+motion_bits)+H

Where num_nozero is the number of non-zero coefficients in the macroblock prediction residual block in the current mode, mode_bits is the number of bits required for the mode information of this mode to be transmitted in the code stream, and motion_bits is the number of bits required for the motion vector to be transmitted in the code stream , η takes 10×QP, and λ takes 0.69×QP, where QP refers to the quantization parameter of the current macroblock. the

Step 2, compare the candidate mode selected in step 1 with the coding cost of the SKIP mode, and the coding cost of the SKIP mode is calculated according to the following method. the

Cost_skip＝α×SAD_Skip+β×skipmode_bits

SAD_Skip is the absolute value sum of macroblock residuals after SKIP mode compensation, skipmode_bits is the number of bits required to transmit the mode information of SKIP mode in the code stream, α is taken as 0.85, and β is taken as 0.62×QP, where QP refers to the current macro The quantization parameter of the block. the

If Cost_skip is smaller than minCost_normal and SAD_Skip<Th(QP), then the final coding mode of the macroblock is SKIP mode, otherwise, the mode selected in step 1 is used as the final coding mode of the current macroblock. the

Example 3:

Taking the H.264 standard as an example, in the H.264 standard, the macroblock mode is divided into two categories: an inter-frame coding mode and an intra-frame coding mode. The inter-frame coding mode can be divided into 16×16, 16×8, 8×16, 8×8, 8×4, 4×8, 4×4 modes according to the size of the sub-block. In addition, the inter-frame also There are special coding modes, such as SKIP mode in P slice, SKIP and DIRECT mode in B, and DIRECT mode is divided into space DIRECT mode and time DIRECT mode. There are two types of intra-frame encoding modes: 16×16 and 4×4. Among them, there are 4 modes in 16×16: vertical mode, horizontal mode, DC mode, and planar mode; there are 9 intra-frame coding modes in 4×4 blocks: horizontal mode, vertical mode, lower right mode, lower left mode, DC mode, Lower left diagonal mode, lower right diagonal mode, right vertical mode, lower horizontal mode, left vertical mode, upper horizontal mode. For each macroblock, it is necessary to select a mode with high coding efficiency among these coding modes for coding. The method for selecting a macroblock mode according to claim 9 is described below by taking mode selection in a B slice in the H.264 standard as an example. the

The specific implementation steps are as follows:

Step 1, use the following cost function to select a mode with the smallest coding cost in the inter mode 16×16, 16×8, 8×16, 8×8, 8×4, 4×8, 4×4 mode . The encoding cost of the selected macroblock mode is recorded as minCost_inter_int. the

Cost_inter_int4＝SATD_int+λ×(mode_bits+motion_bits)

Among them, SATD_int4 is the absolute value sum of the Hada transform of the residual block after 4 times integer pixel precision motion search in the current mode, mode_bits is the number of bits required for this mode to be transmitted in the code stream, and motion_bits is the motion vector in the code stream To transmit the required number of bits, λ is 0.68×QP, where QP refers to the quantization parameter of the current macroblock. the

Step 2, calculate the encoding cost of the mode selected in step 1 after the whole pixel precision motion search

minCost_inter_int＝SATD_int+λ×(mode_bits+motion_bits)

Among them, SATD_int is the absolute value sum after the Hada transform of the residual block of the mode selected in step 1 after the whole-pixel precision motion search, mode_bits is the number of bits required for the mode information of this mode to be transmitted in the code stream, and motion_bits is The motion vector transmits the required number of bits in the code stream, and λ is 0.68×QP, where QP refers to the quantization parameter of the current macroblock. the

Step 3, using the following cost function to select an intra-frame mode with the smallest encoding cost among the nine intra-frame modes. The encoding cost of the selected macroblock mode is denoted as minCost_intra. the

Cost_intra＝SATD_intra+λ×mode_bits+H

Among them, SATD_intra is the absolute value sum of the Hada transform of the residual block after intra prediction in the current mode, mode_bits is the number of bits required for the mode information of this mode to be transmitted in the code stream, and λ is 0.68×QP, where QP refers to is the quantization parameter of the current macroblock, and H is the correction amount of the intra-frame cost, which is taken as 20. the

Step 4, compare the encoding cost of the mode selected in step 1 and step 3, that is, the size of minCost_inter_int and minCost_intra, if minCost_inter_int is larger, set the mode selected in step 3 as a candidate mode, otherwise set the mode in step 1 Set to candidate mode. the

Step 5, compare the coding cost of the candidate mode selected in step 4 with the optimal DIRECT mode, and select the mode with the lower coding cost as a new candidate mode. Let the encoding cost of the mode selected at this time be minCost_normal. The DIRECT mode includes a temporal DIRECT mode and a spatial DIRECT mode, and their coding costs are calculated according to the following method. the

Cost_direct＝SATD_direct+λ×mode_bits

Among them, SATD_direct is the absolute value sum of the Hada transform of the residual block after DIRECT mode prediction, mode_bits is the number of bits that the mode information of DIRECT mode needs to transmit in the code stream, and λ is 0.68×QP, where QP refers to the current The quantization parameter of the macroblock. the

Step 6, compare the candidate mode selected in step 5 with the encoding cost of the SKIP mode, and the encoding cost of the SKIP mode is calculated according to the following method. the

Cost_skip＝α×SAD_Skip+β×skipmode_bits

SAD_Skip is the absolute value sum of macroblock residuals after SKIP mode compensation, skipmode_bits is the number of bits that need to be transmitted in the code stream in SKIP mode, α is taken as 0.8, and β is taken as 0.63×QP, where QP refers to the current macroblock quantization parameters. the

If Cost_skip is smaller than minCost_normal and SAD_Skip<Th(QP), the final coding mode of the macroblock is SKIP mode, otherwise, the mode selected in step 5 is used as the final coding mode of the current macroblock. the

Example 4:

A device for a fast selection method of a macroblock mode, as shown in FIG. 1 . The device is implemented according to the mode selection method described in Embodiment 1, and specifically includes: a first-type inter-frame mode selector with integer pixel precision, a motion estimator with 1/4 pixel precision, an intra-frame mode selector, a mode decider, and a second Class inter-mode selector. the

1) The output terminal of the first-type inter-frame mode selector with integer pixel precision is connected to a 1/4-pixel precision motion estimator, which outputs the optimal coding mode in the first-type inter-frame mode and the coding cost of the macroblock in this coding mode. The first type of inter-frame mode here refers to 16×16, 16×8, 8×16, 8×8 macroblock coding modes. the

2) The input terminal of the 1/4 pixel precision motion estimator is connected to the first type inter-frame mode selector with integer pixel precision, and the output terminal is connected to the mode decision device, which outputs the corresponding mode and the 1/4 pixel precision motion estimation in this encoding mode The coding cost of the macroblock. the

3) The output terminal of the intra-frame mode selector is connected to the mode decision device, which outputs the optimal mode among the five intra-frame modes and the coding cost of the macroblock in this coding mode. the

4) The output terminal of the second-type inter-frame mode selector is connected to the mode decision device, which outputs the optimal coding mode in the second-type inter-frame mode and the macroblock coding cost in this coding mode. The second type of inter-frame mode here refers to SKIP mode and DIRECT mode. the

5) One input terminal of the mode decision device is connected to the 1/4 pixel precision motion estimator, one input terminal is connected to the intra-frame mode selector, and one input terminal is connected to the second-type inter-frame mode selector, and three input modes are output the best mode in . the

Embodiment 5:

A device for a fast selection method of a macroblock mode, as shown in FIG. 2 . The device is implemented according to the mode selection method described in Embodiment 2, and specifically includes a residual absolute value comparator, a first encoding cost calculator, a second encoding cost calculator, and a mode selector. the

1) The output terminal of the residual absolute value comparator is connected to the mode selector, and outputs a signal indicating whether the residual absolute value SAD_Skip is greater than the threshold Th(QP). the

2) The output terminal of the first coding cost calculator is connected to the mode selector, and outputs the optimal mode among other modes except SKIP mode and the coding cost of the macroblock in this mode. the

3) The output terminal of the second coding cost calculator is connected to the mode selector to output the coding cost of the SKIP mode. the

4) The three input terminals of the mode selector are respectively connected to the above three modules, the residual absolute value comparator, the first encoding cost calculator, and the second encoding cost calculator. Output the finalized encoding mode. the

Embodiment 6:

A device for a fast selection method of a macroblock mode, as shown in FIG. 3 . The device is implemented according to the mode selection method described in Embodiment 3, and specifically includes: a first-type inter-frame mode selector with 4 times integer pixel precision, a motion estimator with integer pixel precision, an intra-frame mode selector, a first mode decision device, The second type inter-frame mode selector, the second mode decision unit, and the SKIP mode calculator. the

1) The output terminal of the first-type inter-frame mode selector with 4 times integer pixel precision is connected to the integer-pixel precision motion estimator, and outputs the optimal coding mode in the first-type inter-frame coding mode and the coding cost of the macroblock in this coding mode . The first type of inter-frame mode here refers to 16×16, 16×8, 8×16, 8×8, 8×4, 4×8, and 4×4 macroblock coding modes. the

2) The input terminal of the integer pixel precision motion estimator is connected to the first type inter-frame mode selector with 4 times the integer pixel precision, and the output terminal is connected to the first mode decision device, which outputs the corresponding mode and high-precision motion estimation in this coding mode The coding cost of the macroblock. the

3) The output terminal of the intra-frame mode selector is connected to the first mode decision device, which outputs the optimal mode among the 16×16 and 4×4 intra-frame modes and the coding cost of the macroblock in this coding mode. the

4) The output terminal of the second-type inter-frame mode selector is connected to the first mode decision device, and outputs the optimal coding mode in the second-type inter-frame mode and the macroblock coding cost in this coding mode. The second type of inter-frame mode here refers to a spatial DIRECT mode and a temporal DIRECT mode. the

5) An input terminal of the first mode decision device is connected to an integer pixel precision motion estimator, an input terminal is connected to an intra-frame mode selector, an input terminal is connected to a second-type inter-frame mode selector, and an output terminal is connected to a second mode The decision unit outputs the optimal mode among the three input modes and the encoding cost of the macroblock in the encoding mode. the

6) The output terminal of the SKIP mode calculator is connected to the second mode decision device, and outputs the SKIP mode and the coding cost of the macroblock in the SKIP mode. the

7) One input terminal of the second mode decision device is connected to the first mode decision device, and the other input terminal is connected to the SKIP mode calculator, and the better coding mode among the two input coding modes is output

Claims (5)

1. A selection method of a macroblock mode, characterized in that, the method is: the SKIP mode of the macroblock adopts the method of combining the threshold judgment of the absolute value of the prediction residual block size and the encoding cost function to select, that is, when In SKIP mode, the absolute value of the prediction residual block is less than a threshold determined by the size of the macroblock quantization parameter, and the encoding cost of SKIP is better than the encoding cost of the optimal mode in other modes except SKIP mode. When the current macroblock selects SKIP mode; otherwise, the current macroblock selects the optimal coding mode in other modes except the SKIP mode; the better coding mode here refers to the coding with a smaller coding cost or other criteria combined with the coding cost model. 2. the selection method of macroblock mode according to claim 1, is characterized in that, encoding cost adopts following computing method to obtain: The encoding cost of other modes except SKIP mode is calculated by the following method: Cost＝C1+λ×C2+T1; Among them, C1 is a parameter to measure the similarity between the pixel value of the current macroblock and the predicted pixel value of the macroblock in the current mode. C1 is a non-negative number. The higher the similarity, the smaller the C1. C1 is equal to zero when the pixels of the predicted macroblock in the mode are exactly the same; λ is a non-negative number; C2 is a parameter describing the number of coding bits of the current macroblock except the quantization coefficient; T1 is a non-negative constant; The encoding cost of SKIP mode is calculated by the following method: Cost = α×C3+β×C4+T2; Among them, C3 is a parameter to measure the similarity between the pixel value of the current macroblock and the predicted pixel value of the SKIP mode reference macroblock. C3 is a non-negative number. The higher the similarity, the smaller the C3. When the predicted pixels of the macroblock in the mode are exactly the same, C1 is equal to zero; α, β are non-negative numbers; C4 is a parameter describing the number of coding bits of the current macroblock except the quantization coefficient; T2 is a non-negative constant. 3. A device for the selection method of the macroblock mode according to claim 1, characterized in that it comprises a residual absolute value comparator, a first encoding cost calculator, a second encoding cost calculator, and a mode selector ;in, The output terminal of the residual absolute value comparator is connected to the mode selector, and outputs a signal indicating whether the residual absolute value is greater than the threshold; The output terminal of the first encoding cost calculator is connected to the mode selector, and outputs the optimal mode in other modes except the SKIP mode and the encoding cost of the macroblock in this mode; The output terminal of the second encoding cost calculator is connected to the mode selector, and outputs the encoding cost of the SKIP mode; The three input ends of the mode selector are respectively connected to the residual absolute value comparator, the first encoding cost calculator and the second encoding cost calculator; the finally selected encoding mode is output. the 4. A selection method of a macroblock mode, characterized in that, the selection method of the macroblock mode comprises the following steps: 1) For the first type of inter-frame mode, the pixel-accurate hierarchical selection method is used to select the optimal first-type inter-frame mode, and then the encoding cost of the optimal intra-frame mode and the optimal second-type inter-frame mode Compare and select the mode with better encoding cost; 2) compare the mode in step 1) with the SKIP mode, and select the optimal encoding mode; when selecting, adopt the method of combining the threshold judgment of the absolute value of the prediction residual block and the encoding cost function; The first type of inter-frame mode here refers to the inter-frame coding mode that requires motion search, the second type of inter-frame mode refers to the inter-frame coding mode that does not require motion search, and the better coding mode refers to the coding mode The coding mode selected by the smaller cost or other judging criteria combined with the size of the coding cost. 5. A device for the selection method of the macroblock mode as claimed in claim 4, characterized in that it comprises: low-precision pixel first class inter-frame mode selector, high-precision pixel motion estimator, intra-frame mode selection device, a first mode decision device, a second type of inter-frame mode selector, a SKIP mode calculator, and a second mode decision device; wherein, The output terminal of the low-precision pixel first-type inter-frame mode selector is connected to the high-precision pixel motion estimator, and outputs the optimal coding mode in the first-type inter-frame mode and the coding cost of the macroblock in this coding mode; The input terminal of the high-precision pixel motion estimator is connected to the low-precision pixel first-type inter-frame mode selector, the output terminal is connected to the mode decision device, and the corresponding mode and the coding cost of the macroblock in the coding mode after the high-precision motion estimation are output; The output terminal of the intra-frame mode selector is connected to the first mode decision device, which outputs the optimal intra-frame mode and the coding cost of the macroblock in the coding mode; The output terminal of the second-type inter-frame mode selector is connected to the first mode decision device, and outputs the optimal coding mode in the second-type inter-frame mode and the macroblock coding cost under the coding mode; An input terminal of the first mode decision device is connected to a high-precision pixel motion estimator, an input terminal is connected to an intra-frame mode selector, an input terminal is connected to a second-type inter-frame mode selector, and an output terminal is connected to a second mode decision device. , output the optimal mode among the three input modes and the coding cost of the macroblock in this coding mode; The output terminal of the SKIP mode calculator is connected to the second mode decision device, which outputs the coding cost of the SKIP mode and the macroblock under the SKIP mode; One input terminal of the second mode determiner is connected to the first mode determiner, and the other input terminal is connected to the SKIP mode calculator, which outputs the better encoding mode among the two input encoding modes. the

Images