WO2020207451A1

WO2020207451A1 - H.265 encoding method and apparatus

Info

Publication number: WO2020207451A1
Application number: PCT/CN2020/084093
Authority: WO
Inventors: 张善旭; 陈恒明; 张圣钦; 何德龙
Original assignee: 福州瑞芯微电子股份有限公司
Priority date: 2019-04-11
Filing date: 2020-04-10
Publication date: 2020-10-15

Abstract

An H.265 encoding method and apparatus, the apparatus comprising the following modules: a pre-processing module (120), a rough selection module (130) and a precise comparison module (140). The pre-processing module (120) is used to segment a current frame in an original video (100) into a plurality of CTU blocks; the rough selection module is used to divide each CTU block according to a plurality of partition modes, and segment each CU block therein into one or more corresponding PU blocks; the rough selection module (130) is also used to perform inter frame prediction and intra frame-prediction for each partition mode of each CTU block, and to generate one or more items of prediction information corresponding to each partition mode; the precise comparison module (140) is used to perform cost comparison on prediction information corresponding to each partition mode of each CTU block, and to generate entropy encoding information used for generating a current frame into an H.265 code stream and reconstruction information for generating the current frame into a reconstructed frame. By means of a distributed search means, searching accuracy is improved and hardware resource consumption is reduced.

Description

A H.265 encoding method and device

Technical field

The present invention relates to the field of H.265 coding, in particular to a H.265 coding method and device.

Background technique

H.265 is a new video coding standard developed by ITU-T VCEG after H.264. The H.265 standard revolves around the existing video coding standard H.264, retaining some of the original technologies, while improving some related technologies. The newly added technology is used to improve the relationship between code stream, coding quality, delay and algorithm complexity to achieve optimal settings. Specific research contents include: improving compression efficiency, improving robustness and error recovery capabilities, reducing real-time delay, reducing channel acquisition time and random access delay, and reducing complexity. At present, the existing H.265 algorithm generally has the problems of large hardware resource consumption and low coding efficiency.

Summary of the invention

For this reason, it is necessary to provide a technical solution for H.265 encoding to reduce the hardware resource consumption of the H.265 algorithm.

In order to achieve the above objective, the inventor provides an H.265 encoding device, which includes the following modules: a preprocessing module, a coarse selection module, and an accurate comparison module. The preprocessing module is connected to the coarse selection module. The module is connected to the precise comparison module; where:

The preprocessing module is used to divide a current frame in an original video into multiple CTU blocks;

The coarse selection module is used to divide each CTU block according to multiple division modes, each division mode divides one CTU block into corresponding multiple CU blocks, and divides each CU block into corresponding one or Multiple PU blocks; the coarse selection module is also used to perform inter-frame prediction and intra-frame prediction on each division mode of each CTU block, and generate prediction information corresponding to each division mode;

The precise comparison module is used to compare the cost of prediction information corresponding to each partition mode of each CTU block, select the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode, and According to the selected division mode and its corresponding coding information, the entropy coding information used to generate the H.265 bitstream from the current frame and the reconstruction information for generating the reconstructed frame from the current frame are generated.

The inventor also provides an H.265 encoding method, which is applied to an H.265 encoding device. The device includes the following modules: a preprocessing module, a coarse selection module, and an accurate comparison module. The coarse selection module is connected, and the coarse selection module is connected with the precise comparison module; the method includes the following steps:

The preprocessing module divides a current frame in an original video into multiple CTU blocks;

The coarse selection module divides each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides each CU block into one or more corresponding PU blocks. ; And perform inter-frame prediction and intra-frame prediction on each division mode of each CTU block, and generate a prediction information corresponding to each division mode;

The precise comparison module compares the cost of the prediction information corresponding to each partition mode of each CTU block, selects the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode, and selects The division mode and its corresponding coding information are used to generate entropy coding information for generating an H.265 code stream from the current frame and reconstruction information for generating a reconstructed frame from the current frame.

The inventor also provides an H.265 encoding device, including multiple modules and multiple pipeline steps, each pipeline step includes at least one pipeline stage for executing at least one module, wherein:

The multiple modules include a preprocessing module, a coarse selection module, an accurate comparison module, and an overall control module, and the overall control module is respectively connected to the preprocessing module, the rough selection module, and the precise comparison module;

The multiple pipeline steps include a pretreatment pipeline step, a rough selection pipeline step, and an accurate comparison pipeline step, the rough selection pipeline step is performed after the pretreatment pipeline step, and the precise comparison pipeline step is executed after the rough selection pipeline step;

The preprocessing pipeline step divides a current frame in an original video into multiple CTU blocks through the preprocessing module;

The rough selection pipeline step uses the rough selection module to divide each CTU block according to multiple division modes, and performs coarse selection of inter prediction and coarse selection of intra prediction for each division mode of each CTU block, and generates a and Forecast information corresponding to each division mode;

The precise comparison pipeline step calculates and compares the prediction information corresponding to each division mode of each CTU block through the precise comparison module, and selects a division mode with the smallest cost for each CTU block and the division mode. Corresponding coding information, and according to the selected division mode and its corresponding coding information, generate entropy coding information for generating the H.265 code stream from the current frame and reconstruction information for generating the reconstructed frame from the current frame,

The overall control module is used to control the storage and retrieval of original frame data and reference frame data, and control the preprocessing module, the coarse selection module, and the precise comparison module to sequentially execute the corresponding pipeline steps.

The inventor also provides a H.265 encoding method, which is applied to an H.265 encoding device, the device includes multiple modules and multiple pipeline steps, each pipeline step includes at least one pipeline stage for execution At least one module, of which:

The method includes the following steps:

The rough selection pipeline process uses the rough selection module to divide each CTU block according to multiple division modes, and performs coarse selection of inter prediction and coarse selection of intra prediction for each division mode of each CTU block, and generates one and each Forecast information corresponding to the division mode;

The precise comparison pipeline step calculates and compares the prediction information corresponding to each partition mode of each CTU block through the precise comparison module, and selects the partition mode with the smallest cost for each CTU block and the partition mode corresponding to the partition mode. Encoding information, and according to the selected division mode and its corresponding encoding information, generating entropy encoding information for generating H.265 bitstream from the current frame and reconstruction information for generating reconstructed frames from the current frame,

The overall control module is used to control the storage and retrieval of original frame data and reference frame data, and to control the preprocessing module, the coarse selection module, and the precise comparison module to sequentially execute the corresponding pipeline steps.

Description of the drawings

FIG. 1 is a schematic diagram of an H.265 encoding device related to an embodiment of the present invention;

2 is a schematic diagram of a coarse selection module of an H.265 encoding device according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a rough search process of an H.265 encoding device according to an embodiment of the present invention;

4 is a schematic diagram of the fine search process of the H.265 encoding device according to an embodiment of the present invention;

5 is a schematic diagram of fractional pixel search of an H.265 encoding device according to an embodiment of the present invention;

6-A is a schematic diagram of search prediction performed by an H.265 encoding device according to an embodiment of the present invention;

FIG. 6-B is a schematic diagram of search prediction performed by an H.265 encoding device according to another embodiment of the present invention;

FIG. 7 is a schematic diagram of an accurate comparison module of an H.265 encoding device according to an embodiment of the present invention;

FIG. 8 is a schematic diagram of a layered comparison module of an H.265 encoding device according to an embodiment of the present invention;

FIG. 9 is a flowchart of an H.265 encoding method according to an embodiment of the present invention;

FIG. 10 is a flowchart of a rough search method for H.265 encoding according to an embodiment of the present invention;

FIG. 11 is a flowchart of a fine search method for H.265 encoding according to an embodiment of the present invention;

FIG. 12 is a flowchart of a H.265 coded fractional pixel search method according to an embodiment of the present invention;

FIG. 13 is a schematic diagram of motion vector information around a current CTU block according to an embodiment of the present invention;

14 is a flowchart of an H.265 encoding method related to another embodiment of the present invention;

15 is a schematic diagram of an H.265 encoding device related to another embodiment of the present invention;

Reference signs:

100. Original video;

101. Original image frame;

102. The current frame;

110. Image coding equipment; 120. Preprocessing module; 130. Coarse selection module; 140. Precise comparison module; 150. Entropy coding module; 160. Deblocking filtering module; 170. Sample adaptive biasing module; 180. Post-processing Module

121. Current CTU; 141, coding information; 180, encoded video; 190, code stream; 145, reconstructed frame image;

230. Inter-frame prediction coarse selection module; 211. Coarse search module; 213. Fine search module; 215. Fractional pixel search module;

330. Intra-frame prediction coarse selection module; 231. Reference pixel generation module;

310. Reference frame; 311, down-sampling; 320, down-sampled image; 351, motion vector; 352, minimum cost pixel block; 330, current CTU; 340, down-sampling CTU.

410. Reference frame; 420, current PU position; 421, restore motion vector; 423, fine search motion vector; 430, fine search area; 431, start search position; 433, minimum cost position;

510. Reference frame; 520. Current PU position; 521. Fine search motion vector; 423. Fractional pixel search motion vector; 530. Fractional pixel search area; 531. Start search position; 533. Minimum cost position;

711. Distribution module; 721, first-level calculation Level_calc0; 722, second-level calculation Level_calc1;

723, three-level calculation Level_calc2; 724, four-level calculation Level_calc3;

740. Hierarchical comparison module;

810. Single-stage calculation module; 820, inter-mode cost calculation module; 830, intra-mode cost calculation module; 840, optimization module;

910. Reference frame data loading module; 920. Overall control module.

detailed description

In order to describe in detail the technical content, structural features, achieved objectives and effects of the technical solution, the following detailed description will be given in conjunction with specific embodiments and accompanying drawings.

Please refer to FIG. 1, which is a schematic diagram of an H.265 encoding apparatus according to an embodiment of the present invention. The device is an image encoding device 110. The device may be a chip with image encoding function, or an electronic device containing the above chip, such as a smart mobile device such as a mobile phone, a tablet computer, a personal digital assistant, or a personal digital assistant. Computers, computers for industrial equipment and other electronic equipment. The device includes the following modules: a preprocessing module 120, a coarse selection module 130, and an accurate comparison module 140, the preprocessing module 120 is connected to the coarse selection module 130, the coarse selection module 130 and the precise comparison module 140 Connection; where:

The preprocessing module 120 is used to divide a current frame 102 in an original video 100 into multiple CTU blocks (Coding Tree Unit, coding tree unit). The CTU is a sub-block in the current frame image, and the size can be any of 16x16 sub-blocks, 32x32 sub-blocks, and 64x64 sub-blocks. Specifically, the preprocessing module may obtain an original image frame 101 in the original video 100, and select a current frame 102 from the original image frame 101.

The coarse selection module 130 is configured to divide each CTU block according to multiple division modes, each division mode divides a CTU block into corresponding multiple CU blocks (Coding Unit, coding unit), and divides each of them The CU block is divided into one or more PU blocks (Prediction Unit, prediction unit); the coarse selection module 130 is also used to perform inter-frame prediction and intra-frame prediction for each division mode of each CTU block, and generate A prediction information corresponding to each division mode. The division mode is selected according to actual needs. For example, for a current CTU 121 with a size of 64x64, it can be divided into 4 32x32 sub-blocks; for each 32x32 sub-block, it can be divided into 4 16x16 sub-blocks.

The precise comparison module 140 is configured to compare the prediction information corresponding to each partition mode of each CTU block, and select the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode, And according to the selected division mode and its corresponding coding information, the entropy coding information for generating the H.265 bitstream from the current frame and the reconstruction information for generating the reconstructed frame from the current frame are generated. In this way, the search accuracy is improved through the distributed search, while the details of the reconstructed image are better preserved, and the hardware resource consumption is reduced.

In some embodiments, the device further includes an entropy encoding module 150, which is connected to the precise comparison module 140: the entropy encoding module 150 is configured to divide according to the least costly corresponding to each CTU block The mode and the entropy coding information corresponding to the current frame generated according to the corresponding coding information to generate the H.265 code stream corresponding to the current frame. Specifically, the precise comparison module 140 generates the data required for entropy coding corresponding to the CTU according to the partition mode and prediction mode with the smallest CTU cost, that is, the coding information 141 shown in FIG. 1, the entropy coding module 150 is used to generate an encoded code stream 190 corresponding to the original video according to the data required for entropy encoding corresponding to the CTU. At the same time, the image encoding device 110 will also output the encoded video 180, and a certain image frame of the encoded video 180 is the reconstructed image frame 145.

In some embodiments, the device includes a post-processing module that is connected to the precise comparison module. The post-processing module is used to generate the reconstruction corresponding to the current frame according to the least costly partition mode corresponding to each CTU block and the reconstruction information corresponding to the current frame generated according to the corresponding coding information frame.

Preferably, the post-processing module includes a deblocking filter module 160 and a sample adaptive offset module 170; the deblocking filter module 160 is connected to the sample adaptive offset module 170; the deblocking filter module 160 is used to use The accurate comparison module provides the least costly partition mode and its corresponding coding information to filter the reconstructed frame; the sample adaptive offset module 170 is used to perform SAO calculation on the filtered reconstructed frame, and The calculated data is transmitted to the entropy encoding module 150.

As shown in FIG. 2, the coarse selection module 130 includes an inter-frame prediction coarse selection module 230 and an intra-frame prediction coarse selection module 330, and the inter-frame prediction coarse selection module 230 is respectively connected to the preprocessing module 120 and the precise comparison module 140 , The intra-frame prediction coarse selection module 330 is respectively connected to the pre-processing module 120 and the precise comparison module 140; wherein:

The inter-frame prediction coarse selection module 230 is configured to perform inter-frame prediction on each PU block in each division mode, and select one or more reference frames with a cost less than a preset cost value relative to each PU block The obtained reference information and the motion vector of the selected reference PU block are used as prediction information corresponding to the division mode. Each PU block has its own corresponding motion vector. The motion vector of each PU block is used to obtain prediction information from the reconstructed reference frame. Specifically, the location of the current PU block can be used as the starting point, and the motion vector of each PU block The corresponding motion vector obtains prediction information.

The intra-frame prediction coarse selection module 330 is configured to perform intra-frame prediction on each PU block in each division mode, and select one or more intra-frame prediction directions whose cost is less than a preset cost value relative to each PU block , And use the selected intra prediction direction as the prediction information corresponding to the division mode.

In some embodiments, the inter-frame prediction coarse selection module 230 further includes: a coarse search module 211, a fine search module 213, and a fractional pixel search module 215. The coarse search module 211 is connected to the preprocessing module 120, so The coarse search module 211 is connected to the fine search module 213, and the fine search module is connected to the 213 fractional pixel search module 215.

The coarse search module is used to select a frame from the reference array, select one of its original frame or reconstructed frame as a reference frame, perform down-sampling operations on the reference frame and the current CTU block, and perform down-sampling on the reference frame after down-sampling Find the pixel location with the least cost compared with the down-sampled CTU block, and calculate the coarse search vector of the pixel location relative to the current CTU block.

The reference list is a list storing reference frames, and the reference frame of the current frame may have multiple frames, all of which are indexed through the reference list. A reference frame includes reconstructed frames and original frames. Since the reference frame and the current CTU block are obtained through down-sampling, the coarse search vector calculated by the coarse search module should also be the corresponding down-sampled search vector, that is, the coarse search vector corresponding to the current CTU block needs to be multiplied by the following The sampling magnification (such as 1/4), and the coarse search vector multiplied by the corresponding magnification is transmitted to the next processing module.

As shown in Figure 3, the coarse search module selects one of the original frame or the reconstructed frame as a reference frame, performs down-sampling operations on the reference frame and the current CTU respectively, and then finds and down-sampled the reference frame after down-sampling. The CTU is compared to the least costly pixel location and coarse search vector. Preferably, in this embodiment, the down-sampling scaling ratio of the reference frame and the current CTU are the same. For example, the down-sampled image 320 obtained from the reference frame 310 after down-sampling 311 is to scale the length and width of the reference frame to 1/4, then the down-sampled CTU obtained by the current CTU 330 after down-sampling 331, through the current The length and width of CTU330 are scaled to 1/4. Then the down-sampled CTU340 (B sub-block in Figure 3) is used as a unit, and prediction is performed in the down-sampled image (A sub-block in Figure 3), and the sampled CTU340 and the down-sampled image 320 are calculated in turn. The cost of the sub-block (take each pixel in the A sub-block as the center, take the sub-block with the same size as the B sub-block), find the pixel block with the smallest cost compared with the down-sampled CTU, and record it as the minimum cost pixel block 352 (C sub-block in Figure 3), and record the center pixel position of the current minimum cost pixel block and the coarse search vector. The coarse search vector is the center pixel and minimum cost pixel block 352 of the CTU340 (sub-block B in Figure 3) after downsampling. The vector displacement between the center pixel positions of (C sub-block in FIG. 3) (that is, the motion vector 351 in FIG. 3).

In some embodiments, the intra-frame prediction coarse selection module 330 further includes a reference pixel generation module 231. The reference pixel generating module 231 is used to generate reference pixels using the original pixels of the current frame for each PU block in each division mode, and to predict all intra-frame directions according to the rules of the H.265 protocol according to the reference pixels. Perform prediction to obtain prediction results in each direction, and calculate the distortion cost with the original pixels according to the prediction results in each direction, and sort the cost from small to large to select one or more intra-frame prediction directions with less cost. The coarse selection method of the intra-frame prediction coarse selection module is similar to that of the inter-frame prediction coarse selection module, and will not be repeated here. The difference between the two is that when performing intra-frame prediction, the original frame is down-sampled to obtain the down-sampled image, and the down-sampled CTU is down-sampled from the original frame to obtain the down-sampled image for prediction; while performing inter-frame prediction At this time, the reference frame is down-sampled to obtain the down-sampled image, and the down-sampled CTU is predicted in the down-sampled image obtained by down-sampling the reference frame.

As shown in Figure 6-A and Figure 6-B, according to the H.265 protocol, the reference pixels should be reconstructed pixels, but in the process of hardware implementation, only the original pixels can be obtained at the current time point, and the reconstructed pixels are often not available. Therefore, the method of replacing reconstructed pixels with original pixels is adopted in the present invention. Taking a 4x4 PU sub-block as an example, the black-filled dots in the figure are edge pixels. According to the H.265 protocol, the 4x4 block (the shadow-filled dots in Figure 6-B) has a total of 17 boundary pixels. First, the black filled part of the pixels in the figure (ie, side pixels) should be filled with reconstructed pixels, but the reconstructed pixels cannot be obtained at the current time point, and only original pixels are used instead. The shadow filling part is a PU block of 4x4 size. After the boundary pixel filling is completed, prediction is performed according to the protocol to obtain a 4x4 block filled with the shadow part.

As shown in Figure 4, the fine search module sets a fine search area in the reference frame for each PU according to the coarse search vector, and finds a fine search vector corresponding to the PU with the smallest cost in the fine search area . The fine search step is performed in the reference frame 410. Each current CTU contains multiple PUs, and the fine search is performed by selecting one of these PUs as the current PU in a certain order. Specifically, the current PU position 420 is determined first, and then a fine search area 430 is set in the reference frame for the PU according to the previously obtained coarse search vector (or called the restored motion vector 421). Then, a starting search position 431 corresponding to the current PU position 420 is determined in the fine search area 430 according to the restored motion vector 421. Similar to the search method of coarse search, in the fine search area 430, with the starting search position 431 pixels as the center, the pixels in the starting search position 431 and the fine search area 430 are calculated in turn, and the current PU size is the same. For the cost of the sub-block, find the minimum cost position 433, calculate the motion vector between the current PU position 420 and the minimum cost position 433, and record it as the fine search motion vector 423.

In some embodiments, the fine search module is configured to set a fine search area in the reconstructed image of the reference frame for each PU block according to the coarse search vector, and generate a fine search area in the fine search area. A fine search vector with the lowest cost corresponding to the PU block; and used to generate one or more predicted motion vectors with the same function as the coarse search vector according to the motion vector information around the current CTU block, and generate a fine search based on the predicted motion vector Vector; and send all the generated fine search vectors to the fractional pixel search module.

As shown in Figure 13, for a 64x64 current CTU block, in the upper 10 8x8 size sub-blocks (sub-blocks marked with 1-10 in Figure 13), the adjacent CTU block on the upper left side is the same as the upper right side. In adjacent CTU blocks, there is a corresponding rough search result and corresponding motion vector information. In addition, there are 16 assisted motion vectors in the current CTU block, so there are at most 28 mvs as adjacent mvs (that is, the motion vector information around the current CTU block). The 28 motion vector information will undergo a certain screening, and a preset number (such as 3) of adjacent mvs will be screened out and transmitted to the fine search module to determine the same preset number of fine search motion vectors. In this embodiment, the same function means that the filtered preset number of adjacent mvs are the same as the search results obtained by the coarse search module, that is, they will be input to the interface of the fine search module for further processing.

In this embodiment, the coarse search module will input a motion vector to the fine search module, and then select several mvs from adjacent mvs to input to the fine search module. Assuming that there are a total of N mvs input to the fine search module, then the fine search module The search module will also generate N fine search rmvs (that is, fine search vectors), and input the N fine search vectors to FME (that is, the fractional pixel search module), and then FME will compare the costs from these N fine search mvs An optimal fme_mv (ie, fractional pixel search vector) is obtained, and this fme_mv will finally be input to the accurate comparison module.

As shown in FIG. 5, in order to further improve the search accuracy, the fractional pixel search module 215 is configured to set a corresponding fractional pixel search area 530 in the reference frame for each PU block according to each received fine search vector. , And generate a fractional pixel search vector 423 with the lowest cost corresponding to the PU block in the fractional pixel search area 530. Specifically, the fractional pixel search area 530 can be determined in the following manner: according to the current PU position 520 and the previously acquired fine search motion vector, the start search position 531 corresponding to the current PU position 520 is determined in the reference frame 510 to start the search The position pixel is the center, and K pixels are expanded in 4 directions respectively (the value of K can be set according to actual needs), and a square area with a side length of 2K is obtained as the fractional pixel search area 530. Similar to the search method of the fine search, the starting search position 531 pixel is taken as the center, the starting search position 531 and each pixel point in the fractional pixel search area 530 are calculated in turn, and the current PU size is the same. The minimum cost position 533 is calculated, and the motion vector between the current PU position and the minimum cost position 533 is calculated and recorded as the fractional pixel search motion vector 523.

Please refer to FIG. 7, which is a schematic diagram of the precise comparison module in the H.265 encoding device according to an embodiment of the present invention. In some embodiments, the precise comparison module 140 includes a distribution module 711, multiple single-stage calculation modules (such as 721, 722, 723, and 724), and multiple hierarchical comparison modules 740. The distribution module 711 is connected to the coarse selection module 130 and connected to a plurality of single-stage calculation modules; each single-stage calculation module is connected to a corresponding hierarchical comparison module 740. among them:

The distribution module 711 is configured to distribute different prediction information corresponding to the CU block in each division mode to different single-stage calculation modules according to different division modes of each CTU block;

The single-stage calculation module is used to calculate multiple cost information and compare them in layers according to the prediction information corresponding to the CU block received from the distribution module 711, and select a prediction mode with the least cost corresponding to the CU block and Partition mode

The layered comparison module 740 is used to compare the cost information calculated by the single-stage comparison modules of different layers, and select the partition mode with the smallest cost for the CTU block and the corresponding coding information.

In some embodiments, the exact comparison module 140 of FIG. 7 includes four single-

stage calculation modules

721, 722, 723, and 724. Each single-

stage calculation module

721, 722, 723, and 724 may be composed of the single-stage calculation module 810 of FIG. 8. As shown in FIG. 8, the single-stage calculation module 810 includes an inter-mode cost calculation module 820, an intra-mode cost calculation module 830, and an optimization module 840. For each input CU, the single-stage calculation module 810 may calculate an inter-frame cost through the inter-mode cost calculation module 820, and calculate an intra-frame cost by the intra-mode cost calculation module 830, and compare it by the optimization module 840 For the inter-frame cost and the intra-frame cost, determine the partition mode and prediction mode with the smallest comprehensive cost, that is, the partition mode and prediction mode with the smallest cost corresponding to the currently input CU.

Returning to the embodiment of FIG. 7, each single-

stage calculation module

721, 722, 723, and 724 is used to process a CU block of a specific level. For example, the single-stage calculation module 721 can be set as a first-level calculation module for processing 64x64 CU blocks; the single-stage calculation module 722 can be set as a second-level calculation module for processing CU blocks of 32x32 size; single-stage calculation module 723 can be set as a three-level calculation module for processing 16x16 CU blocks; single-level calculation module 724 can be set as a four-level calculation module for processing 8x8 CU blocks. Assume that the precise comparison module 140 receives a CTU from the coarse selection module 130 and the corresponding division mode, prediction information, and multiple inter-frame motion vectors and reference information. The distribution module 711 can distribute to the computing modules 721-724 at all levels according to the size of the CU in various division modes.

In some embodiments, the intra-mode cost calculation module 830 of each single-stage calculation module receives one or more intra-frame prediction information related to a CU of a certain level, calculates and selects an intra-frame cost . The inter-mode cost calculation module 820 of each single-stage calculation module simultaneously/parallel receives one or more inter-frame motion vectors and reference information related to a CU of a certain level, calculates and selects an inter-frame cost. After that, the optimization module 840 of each single-stage calculation module will select a minimum cost from the calculated intra-frame cost and inter-frame cost. In other words, when the minimum cost is intra-frame cost, it means that it is a better choice to use relevant intra-frame prediction information for H.265 encoding; when the minimum cost is inter-frame cost, it means that the relevant inter-frame motion vector and Reference information for H.265 encoding is a better choice.

For example, the hierarchical comparison module 743 can compare the sum of the minimum costs corresponding to four 8x8 blocks calculated by the four-level calculation module 724 with the minimum cost of one 16x16 block calculated from the three-level calculation module 723, And get the less expensive division mode. Specifically, one of the objects to be compared for hierarchical comparison: 4 8x8 blocks (assumed to be called A, B, C, and D blocks), which can all be the smallest cost blocks obtained by inter-frame comparison, and all are intra-frame comparison The obtained minimum cost block, or both the minimum cost block obtained by inter-frame comparison and the minimum cost block obtained by intra-frame comparison. For example, block A can be acquired between frames, and blocks B, C, and D can be acquired within frames. Or blocks A and C can be obtained between frames, and blocks B and D are obtained within frames.

Similarly, the hierarchical comparison module 742 can select four 16x16 blocks with the lowest cost obtained from the hierarchical comparison module 743, and combine them with one 32x32 block with the lowest cost calculated from the secondary calculation module 722 for comparison. Specifically, the four 16x16 blocks (supposedly called E, F, G, and H blocks) selected by the hierarchical comparison module 742 may include a complete 16x16 CU block, or may be composed of multiple 8x8 blocks. For example, the E block may be a 16x16CU block obtained between frames; the F block may be a 16x16CU block obtained within a frame; and the G blocks may be a 16x16 combined block composed of four 8x8 blocks obtained between frames and intraframes.

Similarly, the hierarchical comparison module 741 can select four 32x32 blocks with the smallest cost obtained from the hierarchical comparison module 742, and combine them with one 64x64 block with the smallest cost calculated from the first-level calculation module 721 for comparison. Specifically, the 4 32x32 blocks selected by the hierarchical comparison module 741 (assumed to be called I, J, K, L blocks) can include a complete 32x32CU block, or can be composed of multiple 16x16 blocks, each of which is composed of A combination block composed of multiple 8x8 blocks. For example, the I block can be a 32x32CU block acquired between frames; the J block is composed of four 16x16CU blocks acquired between frames and intraframe; one or more 16x16 blocks in the K block can be composed of multiple 8x8 blocks. Composed of blocks.

Through the above method, the hierarchical comparison module 740 can find the combination of the CTU, CU, and PU block with the smallest cost, and select the partition mode with the smallest cost for the CTU block and the corresponding coding information.

As shown in Figure 9, the inventor also provides a H.265 encoding method, which is applied to an H.265 encoding device, and the device includes the following modules: a preprocessing module, a coarse selection module, and an accurate comparison module. The preprocessing module is connected to the coarse selection module, and the coarse selection module is connected to the precise comparison module; the method includes the following steps:

First enter step S101, the preprocessing module divides a current frame in an original video into multiple CTU blocks;

Then enter step S102. The coarse selection module divides each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides each CU block into a corresponding one or Multiple PU blocks; and perform inter-frame prediction and intra-frame prediction on each division mode of each CTU block, and generate prediction information corresponding to each division mode;

Then it proceeds to step S103. The precise comparison module performs cost comparison on the prediction information corresponding to each partition mode of each CTU block, selects the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode, and According to the selected division mode and its corresponding coding information, the entropy coding information used to generate the H.265 bitstream from the current frame and the reconstruction information for generating the reconstructed frame from the current frame are generated.

In some embodiments, the device further includes an entropy encoding module connected to an accurate comparison module; the method includes the following steps: the entropy encoding module according to the least costly partition mode corresponding to each CTU block And the entropy coding information corresponding to the current frame generated according to the corresponding coding information to generate the H.265 code stream corresponding to the current frame.

In some embodiments, the device includes a post-processing module that is connected to the precise comparison module: the method includes: the post-processing module according to the least costly division mode and the basis of the corresponding to each CTU block The reconstruction information corresponding to the current frame generated from the corresponding encoding information is used to generate a reconstructed frame corresponding to the current frame.

Preferably, the post-processing module includes a deblocking filter module and a sample adaptive offset module; the deblocking filter module is connected to the sample adaptive offset module; the method includes: the deblocking filter module uses the precise comparison module Provide the least costly partition mode and its corresponding coding information, and filter the reconstructed frame; the sample adaptive offset module performs SAO calculation on the reconstructed frame after the filter processing, and transmits the calculated data to entropy coding Module.

In some embodiments, the coarse selection module includes an inter-frame prediction coarse selection module and an intra-frame prediction coarse selection module. The inter-frame prediction coarse selection module is respectively connected to the preprocessing module and the precise comparison module. The prediction coarse selection module is respectively connected with the preprocessing module and the precise comparison module; the method includes: the inter prediction coarse selection module performs inter prediction on each PU block in each division mode, and selects the relative to each PU block One or more reference information obtained from a reference frame whose cost is less than the preset cost value, and the motion vector of the selected reference PU block is used as the prediction information corresponding to the division mode; the intra-frame prediction coarse selection module performs each division Each PU block in the mode performs intra prediction, and selects one or more intra prediction directions whose cost is less than the preset cost value relative to each PU block, and uses the selected intra prediction direction as the division mode corresponding Forecast information.

In some embodiments, the intra-frame prediction coarse selection module further includes a reference pixel generation module; the method includes: the reference pixel generation module uses the original pixels of the current frame for each PU block in each division mode. Generate reference pixels, and predict all intra-frame prediction directions according to the rules of the H.265 protocol to obtain the prediction results in each direction according to the reference pixels, and calculate the distortion cost with the original pixels according to the prediction results in each direction, and reduce the cost One or more intra-frame prediction directions with a lower cost are selected in a large order.

As shown in FIG. 10, in some embodiments, the inter-frame prediction coarse selection module further includes: a coarse search module, a fine search module, and a fractional pixel search module. The coarse search module is connected to the preprocessing module, so The coarse search module is connected with the fine search module, and the fine search module is connected with the score pixel search module. The method includes:

First, go to step S201, the coarse search module to select a frame from the reference array, and select a reference frame from its original frame or reconstructed frame; then go to step S202 to down-sample the reference frame and the current CTU block; then go to step S203. In the sampled reference frame, find the pixel position with the least cost compared with the down-sampled CTU block, and calculate the coarse search vector of the pixel position relative to the current CTU block.

As shown in Figure 11, in some embodiments, the method includes:

First, go to step S301, the fine search module sets a fine search area in the reconstructed image of the reference frame for each PU block according to the rough search vector; then go to step S302 to generate a corresponding PU block in the fine search area A fine search vector with the least cost; and according to the motion vector information around the current CTU block, one or more predicted motion vectors with the same function as the coarse search vector are generated, and a fine search vector is generated according to the predicted motion vector; All fine search vectors are sent to the fractional pixel search module.

As shown in Figure 12, in some embodiments, the method includes:

First, go to step S401, the fractional pixel search module, according to each received fine search vector, set a corresponding fractional pixel search area in the reference frame for each PU block; then go to step S402 to generate one in the fractional pixel search area A fractional pixel search vector with the smallest cost corresponding to the PU block.

In some embodiments, the precise comparison module includes a distribution module, multiple hierarchical calculation modules, and multiple hierarchical comparison modules, the distribution module is connected to the coarse selection module, and the hierarchical comparison module is connected to the distribution module Connection; the method includes:

The distribution module distributes each CU block in each division mode and the prediction information corresponding to the CU block to different layered calculation modules according to each division mode of each CTU block;

The layered calculation module calculates multiple cost information according to the received prediction information corresponding to the CU block and performs intra-layer comparison, and selects a prediction mode and partition mode with the least cost corresponding to the CU block;

The layered comparison module compares the prediction mode selected by the layered calculation modules of different layers and the minimum cost corresponding to the partition mode, and selects the partition mode with the smallest cost for the CTU block and the corresponding coding information.

According to the H.265 encoding method and device of the above technical solution, the device includes the following modules: a preprocessing module, a coarse selection module, and an accurate comparison module, the preprocessing module is connected to the coarse selection module, the coarse selection module Connected to the precise comparison module; wherein: the preprocessing module is used to divide a current frame in an original video into multiple CTU blocks; the coarse selection module is used to divide each CTU according to multiple division modes Block, each division mode divides a CTU block into corresponding multiple CU blocks, and divides each of the CU blocks into corresponding one or more PU blocks; the coarse selection module is also used for each CTU block Each division mode of the block performs inter-frame prediction and intra-frame prediction, and generates a prediction information corresponding to each division mode; the precise comparison module is used for prediction corresponding to each division mode of each CTU block The information is compared with the cost, the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode are selected, and the selected partition mode and its corresponding coding information are used to generate H The entropy coding information of the .265 bitstream and the reconstruction information for generating the reconstructed frame from the current frame. The invention improves the search accuracy through the distributed search mode, while better retaining the details of the reconstructed image, and reduces the hardware resource consumption.

The H.265 encoding device designed in the present invention can adopt another implementation manner of a pipeline including multiple pipeline steps to implement multiple steps in a specific embodiment. The above-mentioned "pipeline", also known as pipeline (Pipeline), refers to the process of splitting the encoding process of H.265 into multiple steps, and executing these steps in parallel through multiple corresponding hardware processing units to speed up the processing speed. A hardware implementation. "Streamline step" refers to a specific step in a pipeline; "pipeline stage" refers to a specific pipeline stage within a pipeline step. In other words, a pipeline can include one or more pipeline steps; a pipeline step can include one or more pipeline stages. When only one pipeline stage is included in a pipeline step, the pipeline step and pipeline stage can be treated equally.

In some embodiments, a specific hardware module can support the operation of one or more pipeline steps. That is to say, all the pipeline stages in these pipeline steps are run by the hardware module (or by the sub-modules contained therein). In other embodiments, a specific hardware module can support at least one pipeline stage. If there are multiple pipeline stages in a pipeline step, the hardware module is only responsible for the operation of one or more pipeline stages in the pipeline step. In other words, the pipeline step can be implemented by multiple hardware modules, and each hardware module is responsible for running the corresponding pipeline stage in the corresponding pipeline step.

The device includes multiple modules and multiple pipeline steps, each pipeline step includes at least one pipeline stage for executing at least one module, wherein:

The multiple modules include a preprocessing module 120, a coarse selection module 130, an accurate comparison module 140, and an overall control module 920, and the overall control module 920 is connected to the preprocessing module 120, the rough selection module 130, and the precise comparison module 140, respectively;

The preprocessing pipeline step passes through the preprocessing module 120 to divide a current frame 102 in an original video 100 into multiple CTU blocks (Coding Tree Unit, coding tree unit). The CTU is a sub-block in the current frame image, and the size can be any of 16x16 sub-blocks, 32x32 sub-blocks, and 64x64 sub-blocks. Specifically, the preprocessing module may obtain an original image frame 101 in the original video 100, and select a current frame 102 from the original image frame 101.

The coarse selection pipeline step passes through the coarse selection module 130, divides each CTU block according to multiple division modes, performs coarse selection of inter prediction and coarse selection of intra prediction for each division mode of each CTU block, and generates one The prediction information corresponding to each division mode.

As shown in FIG. 2, in this embodiment, the coarse selection module includes: an inter prediction coarse selection module and an intra prediction coarse selection module; the coarse selection pipeline includes: inter prediction coarse selection pipeline and frame Intra-prediction rough selection of pipeline level;

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks (Coding Units, coding units). Unit), and divide each CU block into one or more corresponding PU blocks (Prediction Unit, prediction unit), perform inter-frame prediction for each division mode of each CTU block and obtain reference frame information, and Perform intra-frame prediction on each division mode of each CTU block and generate prediction information corresponding to each division mode. The division mode is selected according to actual needs. For example, for a current CTU 121 with a size of 64x64, it can be divided into 4 32x32 sub-blocks; for each 32x32 sub-block, it can be divided into 4 16x16 sub-blocks.

The intra-frame prediction coarse selection pipeline passes through the intra-frame prediction coarse selection module: performs intra-frame prediction on each PU block in each division mode and calculates the corresponding cost, and selects one or more costs relative to each PU block according to the cost. Intra prediction directions, and the selected intra prediction direction is used as the prediction information corresponding to the division mode. Each PU block has its own corresponding motion vector. The motion vector of each PU block is used to obtain prediction information from the reconstructed reference frame. Specifically, the location of the current PU block can be used as the starting point, and the motion vector of each PU block The corresponding motion vector obtains prediction information.

The precise comparison pipeline step uses the precise comparison module 140 to calculate and compare the prediction information corresponding to each partition mode of each CTU block, and select the partition mode with the smallest cost for each CTU block and compare it with the partition mode. The coding information corresponding to the mode, and according to the selected division mode and its corresponding coding information, the entropy coding information for generating the H.265 bitstream from the current frame and the reconstruction information for generating the reconstructed frame from the current frame are generated. In this way, the search accuracy is improved through the distributed search, while the details of the reconstructed image are better preserved, and the hardware resource consumption is reduced.

The overall control module is used to control the storage and retrieval of original frame data and reference frame data, and to control the preprocessing module, coarse selection module, and precise comparison module to sequentially execute the corresponding pipeline steps. Preferably, the rough selection pipeline step is performed after the pretreatment pipeline step, and the precise comparison pipeline step is executed after the rough selection pipeline step. In short, when the coarse selection module executes the rough selection pipeline corresponding to the current frame, the preprocessing module can perform the preprocessing pipeline steps of the next frame corresponding to the current frame, and the precise comparison module executes the accurate comparison pipeline corresponding to the current frame. When the current frame corresponds to the next frame, the rough selection module can perform the rough selection pipeline steps of the next frame, and so on to achieve pipeline operation, thereby effectively improving the coding efficiency.

In some embodiments, the coarse selection module 130 further includes an inter-frame coarse selection module 230, and the precise comparison module 140 further includes an intra-frame coarse selection module 330; the coarse selection pipeline includes an inter-frame coarse selection pipeline. , The precise comparison pipeline step includes coarse selection of pipeline stages within a frame.

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides the Each CU block is divided into one or more corresponding PU blocks, inter-frame prediction is performed for each division mode of each CTU block and reference frame information is obtained, and each division mode of each CTU block is intra-frame prediction And generate a prediction information corresponding to each division mode;

The intra-frame prediction coarse selection pipeline passes through the intra-frame prediction coarse selection module: performs intra-frame prediction on each PU block in each division mode and calculates the corresponding cost, and selects one or more costs relative to each PU block according to the cost. Intra prediction directions, and the selected intra prediction direction is used as the prediction information corresponding to the division mode.

In short, in the actual application process, the intra-frame coarse selection module 330 can be attached to the coarse selection module 130 or the precise comparison module 140, thereby broadening the application scenarios of the device.

In some embodiments, the inter-frame prediction coarse selection module 230 includes: a coarse search module 211, a reference frame data loading module 910, a fine search module 213, and a fractional pixel search module 215. The rough selection pipeline includes: rough search pipeline, reference frame data loading pipeline, fine search pipeline and fractional pixel search pipeline;

The coarse search pipeline stage passes through the coarse search module: select a frame from the reference array, select a reference frame from its original frame or reconstructed frame, perform down-sampling operations on the reference frame and the current CTU block, and perform down-sampling on the Find the pixel location with the least cost compared with the down-sampled CTU block in the reference frame, and calculate the coarse search vector of the pixel location relative to the current CTU block;

The reference frame data loading pipeline stage is through the reference frame data loading pipeline stage: the coarse search vector of the coarse search pipeline is obtained through the overall control module, and one or more predictions with the same function as the coarse search are obtained according to the motion vector around the CTU block Motion vector, load reference frame data according to the coarse search vector and one or more prediction vectors, and pass it to the fine search pipeline through the overall control module;

The fine search pipeline passes the fine search module: according to the coarse search vector, a fine search area is set in the reconstructed image of the reference frame for each PU block, and a corresponding PU block is generated in the fine search area A fine search vector with the smallest cost; and used to generate one or more predicted motion vectors with the same function as the coarse search vector based on the motion vector information around the current CTU block, and generate a fine search vector based on the predicted motion vector; and Send all the generated fine search vectors to the fractional pixel search module;

The fractional pixel search pipeline level passes through the fractional pixel search module: according to each received fine search vector, a corresponding fractional pixel search area is set in the reference frame for each PU block, and in the fractional pixel search area Generate a fractional pixel search vector with the smallest cost corresponding to the PU block. Preferably, the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are the same pipeline stage, and the intra-frame prediction coarse selection module and the fractional pixel search module are executed in parallel at the same pipeline stage.

In some embodiments, the intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed in the intra-frame prediction coarse selection pipeline; the intra-frame prediction coarse selection pipeline includes: for each division mode Each PU block uses the original pixels of the current frame to generate reference pixels. According to the reference pixels, all intra-frame prediction directions are predicted according to the rules of the H.265 protocol to obtain the prediction results in each direction. The prediction results in each direction are compared with the original The pixel calculates the distortion cost, and sorts the cost from small to large to select one or more intra prediction directions with a small cost.

In some embodiments, the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are different pipeline stages, and the intra-frame prediction coarse selection module is executed at the pipeline stage after the fractional pixel search module. The intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed in the intra-frame prediction coarse selection pipeline; the reference pixel generation module is used for each PU block in each division mode, using the reconstruction of the current frame Pixels to generate reference pixels, predict all intra-frame prediction directions according to the rules of the H.265 protocol according to the reference pixels to obtain the prediction results in each direction, and calculate the distortion cost with the original pixels according to the prediction results in each direction, and reduce the cost One or more intra-frame prediction directions with a lower cost are selected in a large order.

stage calculation modules

721, 722, 723, and 724. Each single-

stage calculation module

Returning to the embodiment of FIG. 7, each single-

stage calculation module

In some embodiments, the intra-frame prediction coarse selection module 330 includes a reference pixel generation module 231; the intra-frame prediction coarse selection module 330 is executed in the intra-frame prediction coarse selection pipeline;

The reference pixel generating module 231 is used to generate reference pixels using the original pixels of the current frame for each PU block in each division mode, and to predict all intra-frame directions according to the rules of the H.265 protocol according to the reference pixels. Perform prediction to obtain prediction results in each direction, and calculate the distortion cost with the original pixels according to the prediction results in each direction, and sort the cost from small to large to select one or more intra-frame prediction directions with less cost.

The coarse selection method of the intra-frame prediction coarse selection module is similar to that of the inter-frame prediction coarse selection module, and will not be repeated here. The difference between the two is that when performing intra-frame prediction, the original frame is down-sampled to obtain the down-sampled image, and the down-sampled CTU is down-sampled from the original frame to obtain the down-sampled image for prediction; while performing inter-frame prediction At this time, the reference frame is down-sampled to obtain the down-sampled image, and the down-sampled CTU is predicted in the down-sampled image obtained by down-sampling the reference frame.

In some embodiments, the device further includes a post-processing module 180, which is connected to the precise comparison module 140; the post-processing module 180 is executed in post-processing pipeline steps, and the post-processing pipeline steps include : Generate a reconstructed frame corresponding to the current frame according to the least costly partition mode corresponding to each CTU block output by the precise comparison module and according to the corresponding reconstruction information.

Preferably, the post-processing module 180 includes a deblocking filtering module 160 and a sample adaptive offset module 170; the post-processing pipeline step includes a deblocking filtering pipeline step and a sample adaptive offset step; the deblocking filtering module Executed in the deblocking filtering pipeline step, the sample adaptive offset step is executed in the sample adaptive offset step; the deblocking filtering pipeline step includes: using the least costly partition mode provided by the accurate comparison module and its corresponding The reconstructed information is reconstructed and the reconstructed frame is filtered; the sample adaptive offset pipeline step includes: performing SAO calculation on the reconstructed frame after the filtering process to obtain the final reconstructed frame for reference and display. The deblocking filtering pipeline step and the sample adaptive offset pipeline step are sequentially executed in the post-processing pipeline stage in sequence.

In some embodiments, the device further includes an entropy encoding module 150 connected to the precise comparison module 140. The entropy encoding module 150 is executed in the entropy encoding pipeline step, and the entropy encoding pipeline step includes: according to the least costly partition mode corresponding to each CTU block output by the precise comparison module 140 and the and generated according to the corresponding encoding information Entropy coding information corresponding to the current frame is used to generate an H.265 code stream corresponding to the current frame. The entropy coding pipeline step and the post-processing pipeline step are executed in parallel at the same pipeline stage.

Specifically, the precise comparison module 140 generates the data required for entropy coding corresponding to the CTU according to the partition mode and prediction mode with the smallest CTU cost, that is, the coding information 141 shown in FIG. 1, the entropy coding module 150 is used to generate an encoded code stream 190 corresponding to the original video according to the data required for entropy encoding corresponding to the CTU. At the same time, the image encoding device 110 will also output the encoded video 180, and a certain image frame of the encoded video 180 is the reconstructed image frame 145.

As shown in Figure 14, the inventor also provides a H.265 encoding method, the method is applied to H.265 encoding device, the device includes multiple modules and multiple pipeline steps, each pipeline step includes at least one The pipeline stage is used to execute at least one module, where:

The method includes the following steps:

First, go to step S101' preprocessing pipeline step to divide a current frame in an original video into multiple CTU blocks through the preprocessing module;

Then enter step S102' rough selection pipeline step through the rough selection module, divide each CTU block according to multiple division modes, and perform coarse selection of inter prediction and coarse selection of intra prediction for each division mode of each CTU block, and Generate a prediction information corresponding to each division mode;

Then enter the step S103' accurate comparison pipeline step, through the accurate comparison module, the prediction information corresponding to each division mode of each CTU block is calculated and compared, and the division mode with the smallest cost for each CTU block is selected and compared The coding information corresponding to the division mode, and according to the selected division mode and its corresponding coding information, the entropy coding information for generating the H.265 code stream from the current frame and the reconstruction information for generating the reconstructed frame from the current frame are generated ,

As shown in FIG. 10, the coarse selection module includes: an inter prediction coarse selection module and an intra prediction coarse selection module; the coarse selection pipeline includes: an inter prediction coarse selection pipeline and an intra prediction coarse selection pipeline ；

The method also includes:

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides each of them The CU block is divided into one or more corresponding PU blocks, and each division mode of each CTU block is inter-predicted and reference frame information is obtained, and each division mode of each CTU block is intra-predicted and generated A prediction information corresponding to each division mode;

Intra-frame prediction coarse selection pipeline through the intra-frame prediction coarse selection module: Perform intra-frame prediction for each PU block in each division mode and calculate the corresponding cost, and select one or more frames relative to the cost of each PU block according to the cost Intra prediction direction, and the selected intra prediction direction is used as the prediction information corresponding to the division mode.

In some embodiments, the coarse selection module further includes a coarse inter-frame selection module, and the precise comparison module further includes a coarse intra-frame selection module; the coarse selection pipeline includes the inter-frame coarse selection pipeline, and the precise comparison module The comparison pipeline step includes coarse selection of pipeline stages within the frame.

The method includes:

In short, the intra-frame coarse selection module can be a part of the coarse selection module or a part of the precise comparison module, thereby effectively broadening the application scenarios of the present invention.

In some embodiments, the inter-frame prediction coarse selection module includes: a coarse search module, a reference frame data loading module, a fine search module, and a fractional pixel search module;

The rough selection pipeline includes: rough search pipeline, reference frame data loading pipeline, fine search pipeline and fractional pixel search pipeline;

The method includes:

As shown in Fig. 10, the rough search pipeline stage passes through the rough search module: first enter step S201, the rough search module selects a frame from the reference array, and selects a reference frame from its original frame or reconstructed frame; then enters step S202 for reference The frame and the current CTU block perform down-sampling operation; then go to step S203 to find the pixel position with the least cost compared with the down-sampled CTU block in the down-sampled reference frame, and calculate the coarse search of the pixel position relative to the current CTU block Vector.

The reference frame data loading pipeline stage The reference frame data loading pipeline stage: obtain the coarse search vector of the coarse search pipeline through the overall control module and obtain one or more predicted motion vectors with the same function as the coarse search according to the motion vectors around the CTU block , Load the reference frame data according to the coarse search vector and one or more prediction vectors, and pass it to the fine search pipeline through the overall control module;

As shown in Figure 11, the fine search pipeline level passes through the fine search module: first enter step S301 according to the coarse search vector, set a fine search area in the reconstructed image of the reference frame for each PU block; then enter step S302, Generate a fine search vector corresponding to the PU block in the fine search area; and generate one or more predicted motion vectors with the same function as the coarse search vector according to the motion vector information around the current CTU block, and Predict the motion vector to generate a fine search vector; and send all the generated fine search vectors to the fractional pixel search module;

As shown in Figure 12, the fractional pixel search pipeline passes through the fractional pixel search module: first enter step S401, the fractional pixel search module sets a corresponding frame in the reference frame for each PU block according to each received fine search vector Then, proceed to step S402 to generate a fractional pixel search vector corresponding to the PU block with the smallest cost in the fractional pixel search area.

In some embodiments, the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are the same pipeline stage, and the intra-frame prediction coarse selection module and the fractional pixel search module are executed in parallel at the same pipeline stage . In short, the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline can be executed in parallel, that is, synchronously, or in sequential order, that is, the intra-frame prediction coarse selection pipeline is executed first, and then executed The fractional pixel search pipeline level.

In some embodiments, the intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed at the intra-frame prediction coarse selection pipeline; the method includes:

The rough selection pipeline of intra prediction includes: for each PU block in each division mode, the original pixels of the current frame are used to generate reference pixels, and all intra prediction directions are performed according to the rules of the H.265 protocol according to the reference pixels. The prediction results in each direction are obtained by prediction, and the distortion cost is calculated with the original pixels according to the prediction results in each direction, and the cost is sorted from small to large to select one or more intra prediction directions with a small cost.

In some embodiments, the multiple modules further include a post-processing module, and the multiple pipeline steps further include a post-processing pipeline step, and the method includes: the post-processing pipeline step passes through the post-processing module and outputs the precise comparison module Each CTU block corresponds to a partition mode with the least cost and a reconstructed frame corresponding to the current frame is generated according to the corresponding reconstruction information. In other embodiments, the multiple modules further include an entropy encoding module, and the multiple pipeline steps further include an entropy encoding pipeline step; the method includes: the entropy encoding pipeline step outputs the precise comparison module through the entropy encoding module Each CTU block corresponds to the least costly partition mode and according to its corresponding entropy coding information, a binary code stream conforming to the H.265 protocol specification is generated.

As shown in Fig. 15, the preprocessing module 120 belongs to the first-level pipeline and executes the preprocessing pipeline steps. The coarse selection module performs rough selection pipeline steps. The coarse selection module includes a coarse search module 211, a reference frame data loading module 910, a fine search module 213, and a fractional pixel search module 215. Correspondingly, the coarse selection pipeline includes a coarse search pipeline (i.e., two-stage pipeline), a reference frame data loading pipeline (i.e., a three-stage pipeline), a fine search pipeline (i.e., a four-stage pipeline), and a fractional pixel search pipeline ( That is five-level pipeline). Preferably, the intra-frame prediction coarse selection module and the fractional pixel search module are executed in parallel at the same pipeline stage (that is, both are executed in a five-stage pipeline). The precise comparison module 140 executes the precise comparison pipeline, which belongs to the six-stage pipeline. The entropy coding module 150 and the post-processing module are respectively executed in the entropy coding pipeline stage and the post-processing pipeline stage, and the entropy coding pipeline stage and the post-processing pipeline stage are executed in parallel in the seven-stage pipeline. One to seven levels of pipelines are all implemented through the overall control module 920 to achieve data transmission, scheduling, and control, so that the coding process is carried out in an orderly manner, which greatly improves the coding efficiency.

The present invention provides a H.265 encoding method and device. The device includes multiple modules and multiple pipeline steps. Each pipeline step includes at least one pipeline stage for executing at least one module. The multiple modules include preprocessing. Module, rough selection module, accurate comparison module and overall control module; multiple pipeline steps include pretreatment pipeline step, rough selection pipeline step, and accurate comparison pipeline step. The rough selection pipeline step is executed after the pretreatment pipeline step, so The precise comparison pipeline step is performed after the rough selection pipeline step. The overall control module is used to control the storage and retrieval of original frame data and reference frame data, and to control the preprocessing module, the coarse selection module, and the precise comparison module to sequentially execute the corresponding pipeline steps. The invention improves the search accuracy through the distributed search mode, while better retaining the details of the reconstructed image, and reduces the hardware resource consumption.

It should be noted that although the foregoing embodiments have been described in this article, the scope of patent protection of the present invention is not limited thereby. Therefore, based on the innovative concept of the present invention, changes and modifications to the embodiments described herein, or equivalent structures or equivalent process transformations made by using the description and drawings of the present invention, directly or indirectly apply the above technical solutions In other related technical fields, they are all included in the scope of patent protection of the present invention.

Claims

An H.265 encoding device, which is characterized by comprising the following modules: a preprocessing module, a coarse selection module, and an accurate comparison module, the preprocessing module is connected to the coarse selection module, and the coarse selection module is connected to the precise comparison module. Compare module connection; among them:

The preprocessing module is used to divide a current frame in an original video into multiple CTU blocks;

The coarse selection module is used to divide each CTU block according to multiple division modes, each division mode divides one CTU block into corresponding multiple CU blocks, and divides each CU block into corresponding one or Multiple PU blocks; the coarse selection module is also used to perform inter-frame prediction and intra-frame prediction on each division mode of each CTU block, and generate prediction information corresponding to each division mode;

The precise comparison module is used to compare the cost of prediction information corresponding to each partition mode of each CTU block, select the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode, and According to the selected division mode and its corresponding coding information, the entropy coding information used to generate the H.265 bitstream from the current frame and the reconstruction information for generating the reconstructed frame from the current frame are generated.
The H.265 encoding device according to claim 1, further comprising an entropy encoding module, and the entropy encoding module is connected to the precise comparison module:

The entropy coding module is used to generate H.265 corresponding to the current frame according to the partition mode with the lowest cost corresponding to each CTU block and the entropy coding information corresponding to the current frame generated according to the corresponding coding information. Code stream.
The H.265 encoding device according to claim 2, characterized in that it comprises a post-processing module which is connected to the precise comparison module:

The post-processing module is used to generate the reconstruction corresponding to the current frame according to the least costly partition mode corresponding to each CTU block and the reconstruction information corresponding to the current frame generated according to the corresponding coding information frame.
The H.265 encoding device according to claim 3, wherein the post-processing module comprises a deblocking filtering module and a sample adaptive offset module; the deblocking filtering module is connected to the sample adaptive offset module;

The deblocking filtering module is used for filtering the reconstructed frame by using the partition mode with the least cost provided by the accurate comparison module and the corresponding coding information;

The sample adaptive offset module is used to perform SAO calculation on the reconstructed frame after filtering processing, and transmit the calculated data to the entropy coding module.
The H.265 encoding device according to claim 1, wherein the coarse selection module comprises an inter-frame prediction coarse selection module and an intra-frame prediction coarse selection module, the inter-frame prediction coarse selection module and the preprocessing module respectively 1. The precise comparison module is connected, and the intra-frame prediction coarse selection module is respectively connected with the preprocessing module and the precise comparison module; wherein:

The inter-frame prediction coarse selection module is used to perform inter-frame prediction on each PU block in each division mode, and select one or more reference frames with a cost less than a preset cost value relative to each PU block Reference information of, and the motion vector of the selected reference PU block as the prediction information corresponding to the division mode;

The intra-frame prediction coarse selection module is used to perform intra-frame prediction on each PU block in each division mode, and select one or more intra-frame prediction directions with a cost less than a preset cost value relative to each PU block, And use the selected intra prediction direction as the prediction information corresponding to the division mode.
The H.265 encoding device according to claim 5, wherein the intra-frame prediction coarse selection module further comprises a reference pixel generation module;

The reference pixel generation module is used to generate reference pixels using the original pixels of the current frame for each PU block in each division mode, and perform all intra prediction directions according to the rules of the H.265 protocol according to the reference pixels. The prediction results in each direction are obtained by prediction, and the distortion cost is calculated with the original pixels according to the prediction results in each direction, and the cost is sorted from small to large to select one or more intra prediction directions with a small cost.
The H.265 encoding device according to claim 5, wherein the inter-frame prediction coarse selection module further comprises: a coarse search module, a fine search module, and a fractional pixel search module, the coarse search module and preprocessing Module connection, the coarse search module is connected with the fine search module, and the fine search module is connected with the fractional pixel search module.
The H.265 encoding device according to claim 7, wherein:

The coarse search module is used to select a frame from the reference array, select a reference frame in its original frame or reconstructed frame, perform down-sampling operations on the reference frame and the current CTU block, and find the down-sampled reference frame The pixel location with the least cost compared with the down-sampled CTU block, and the coarse search vector of the pixel location relative to the current CTU block is calculated.
The H.265 encoding device according to claim 7, wherein:

The fine search module is used to set a fine search area in the reconstructed image of the reference frame for each PU block according to the coarse search vector, and generate a PU block corresponding to the smallest cost in the fine search area A fine search vector; and used to generate one or more predicted motion vectors with the same function as the coarse search vector based on the motion vector information around the current CTU block, and generate a fine search vector based on the predicted motion vector; and use all the generated The refined search vector is sent to the fractional pixel search module.
The H.265 encoding device according to claim 9, wherein:

The fractional pixel search module is used to set a corresponding fractional pixel search area in the reference frame for each PU block according to each received fine search vector, and generate a PU block in the fractional pixel search area Corresponding to a fractional pixel search vector with the smallest cost.
The H.265 encoding device according to claim 1, wherein the precise comparison module includes a distribution module, multiple hierarchical calculation modules, and multiple hierarchical comparison modules, and the distribution module is connected to the coarse selection module , The layered comparison module is connected to the distribution module, wherein:

The distribution module is configured to distribute each CU block in each division mode and the prediction information corresponding to the CU block to different hierarchical calculation modules according to each division mode of each CTU block;

The layered calculation module is configured to calculate multiple cost information according to the received prediction information corresponding to the CU block and perform intra-layer comparison, and select a prediction mode and division mode with the least cost corresponding to the CU block;

The layered comparison module is used to compare the prediction mode selected by the layered calculation modules of different layers and the minimum cost corresponding to the partition mode, and select the partition mode with the smallest cost for the CTU block and the corresponding coding information.
An H.265 encoding method, characterized in that the method is applied to an H.265 encoding device, and the device includes the following modules: a preprocessing module, a coarse selection module, and an accurate comparison module, the preprocessing module and the The coarse selection module is connected, and the coarse selection module is connected with the precise comparison module; the method includes the following steps:

The preprocessing module divides a current frame in an original video into multiple CTU blocks;

The coarse selection module divides each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides each CU block into one or more corresponding PU blocks. ; And perform inter-frame prediction and intra-frame prediction on each division mode of each CTU block, and generate a prediction information corresponding to each division mode;

The precise comparison module compares the cost of the prediction information corresponding to each partition mode of each CTU block, selects the partition mode with the smallest cost for each CTU block and the coding information corresponding to the partition mode, and selects The division mode and its corresponding coding information are used to generate entropy coding information for generating an H.265 code stream from the current frame and reconstruction information for generating a reconstructed frame from the current frame.
The H.265 encoding method according to claim 12, wherein the device further comprises an entropy encoding module, and the entropy encoding module is connected to an accurate comparison module; the method comprises the following steps:

The entropy coding module generates the H.265 code stream corresponding to the current frame according to the partition mode with the lowest cost corresponding to each CTU block and the entropy coding information corresponding to the current frame generated according to the corresponding coding information.
The H.265 encoding method according to claim 13, wherein the device comprises a post-processing module, and the post-processing module is connected to an accurate comparison module: the method comprises:

The post-processing module generates a reconstructed frame corresponding to the current frame according to the partition mode with the lowest cost corresponding to each CTU block and the reconstruction information corresponding to the current frame generated according to the corresponding coding information.
The H.265 encoding method according to claim 14, wherein the post-processing module comprises a deblocking filtering module and a sample adaptive offset module; the deblocking filtering module is connected to the sample adaptive offset module; The method includes:

The deblocking filtering module uses the least costly partition mode provided by the accurate comparison module and the corresponding coding information to filter the reconstructed frame;

The sample adaptive offset module performs SAO calculation on the reconstructed frame after the filtering process, and transmits the calculated data to the entropy coding module.
The H.265 encoding method according to claim 12, wherein the coarse selection module comprises an inter-frame prediction coarse selection module and an intra-frame prediction coarse selection module, and the inter-frame prediction coarse selection module and the preprocessing module are respectively , The precise comparison module is connected, the intra-frame prediction coarse selection module is respectively connected with the preprocessing module and the precise comparison module; the method includes:

The inter-frame prediction coarse selection module performs inter-frame prediction on each PU block in each division mode, and selects one or more reference information obtained from the reference frame whose cost is less than the preset cost value relative to each PU block, And use the motion vector of the selected reference PU block as the prediction information corresponding to the division mode;

The intra-frame prediction coarse selection module performs intra-frame prediction on each PU block in each division mode, and selects one or more intra-frame prediction directions whose cost is less than the preset cost value relative to each PU block, and selects The intra prediction direction is used as the prediction information corresponding to the division mode.
The H.265 encoding method according to claim 16, wherein the intra-frame prediction coarse selection module further comprises a reference pixel generation module; the method comprises:

The reference pixel generation module uses the original pixels of the current frame to generate reference pixels for each PU block in each division mode, and predicts all intra-frame prediction directions according to the rules of the H.265 protocol according to the reference pixels to obtain each direction According to the prediction results of each direction, the distortion cost is calculated with the original pixels separately, and the cost is sorted from small to large to select one or more intra prediction directions with a small cost.
The H.265 encoding method according to claim 16, wherein the inter-frame prediction coarse selection module further comprises: a coarse search module, a fine search module, and a fractional pixel search module, the coarse search module and the preprocessing Module connection, the coarse search module is connected with the fine search module, and the fine search module is connected with the fractional pixel search module.
The H.265 encoding method of claim 18, wherein the method comprises:

The coarse search module is used to select a frame from the reference array, select a reference frame in its original frame or reconstructed frame, perform down-sampling operations on the reference frame and the current CTU block, and find and down-sample the reference frame after down-sampling. The sampled CTU block compares the pixel location with the least cost, and calculates the coarse search vector of the pixel location relative to the current CTU block.
The H.265 encoding method of claim 18, wherein the method comprises:

According to the coarse search vector, the fine search module sets a fine search area in the reconstructed image of the reference frame for each PU block, and generates a fine search vector corresponding to the PU block in the fine search area with the smallest cost. ; And according to the motion vector information around the current CTU block, generate one or more predicted motion vectors with the same function as the coarse search vector, and generate the fine search vector based on the predicted motion vector; and send all the generated fine search vectors to the score Pixel search module.
The H.265 encoding method of claim 20, wherein the method comprises:

The fractional pixel search module sets a corresponding fractional pixel search area in the reference frame for each PU block according to each received fine search vector, and generates a corresponding PU block in the fractional pixel search area with the least cost A fractional pixel search vector.
The H.265 encoding method according to claim 12, wherein the precise comparison module includes a distribution module, multiple layered calculation modules, and multiple layered comparison modules, and the distribution module is connected to the coarse selection module , The layered comparison module is connected to the distribution module; the method includes:

The distribution module distributes each CU block in each division mode and the prediction information corresponding to the CU block to different layered calculation modules according to each division mode of each CTU block;

The layered calculation module calculates multiple cost information according to the received prediction information corresponding to the CU block and performs intra-layer comparison, and selects a prediction mode and partition mode with the least cost corresponding to the CU block;

The layered comparison module compares the prediction mode selected by the layered calculation modules of different layers and the minimum cost corresponding to the partition mode, and selects the partition mode with the smallest cost for the CTU block and the corresponding coding information.
An H.265 encoding device, which is characterized by comprising multiple modules and multiple pipeline steps, each pipeline step includes at least one pipeline stage for executing at least one module, wherein:

The multiple modules include a preprocessing module, a coarse selection module, an accurate comparison module, and an overall control module, and the overall control module is respectively connected to the preprocessing module, the rough selection module, and the precise comparison module;

The multiple pipeline steps include a pretreatment pipeline step, a rough selection pipeline step, and an accurate comparison pipeline step, the rough selection pipeline step is performed after the pretreatment pipeline step, and the precise comparison pipeline step is executed after the rough selection pipeline step;

The preprocessing pipeline step divides a current frame in an original video into multiple CTU blocks through the preprocessing module;

The rough selection pipeline step uses the rough selection module to divide each CTU block according to multiple division modes, and performs coarse selection of inter prediction and coarse selection of intra prediction for each division mode of each CTU block, and generates a and Forecast information corresponding to each division mode;

The precise comparison pipeline step calculates and compares the prediction information corresponding to each division mode of each CTU block through the precise comparison module, and selects a division mode with the smallest cost for each CTU block and the division mode. Corresponding coding information, and according to the selected division mode and its corresponding coding information, generate entropy coding information for generating the H.265 code stream from the current frame and reconstruction information for generating the reconstructed frame from the current frame,

The overall control module is used to control the storage and retrieval of original frame data and reference frame data, and control the preprocessing module, the coarse selection module, and the precise comparison module to sequentially execute the corresponding pipeline steps.
The H.265 encoding device according to claim 23, wherein:

The coarse selection module includes: an inter prediction coarse selection module and an intra prediction coarse selection module; the coarse selection pipeline includes: an inter prediction coarse selection pipeline and an intra prediction coarse selection pipeline;

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides the Each CU block is divided into one or more corresponding PU blocks, inter-frame prediction is performed for each division mode of each CTU block and reference frame information is obtained, and each division mode of each CTU block is intra-frame prediction And generate a prediction information corresponding to each division mode;

The intra-frame prediction coarse selection pipeline passes through the intra-frame prediction coarse selection module: performs intra-frame prediction on each PU block in each division mode and calculates the corresponding cost, and selects one or more costs relative to each PU block according to the cost. Intra prediction directions, and the selected intra prediction direction is used as the prediction information corresponding to the division mode.
The H.265 encoding device according to claim 23, wherein:

The coarse selection module further includes a coarse inter-frame selection module, and the precise comparison module further includes a coarse intra-frame selection module;

The rough selection pipeline includes a rough selection pipeline between frames, and the accurate comparison pipeline includes a rough selection pipeline within a frame;

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides the Each CU block is divided into one or more corresponding PU blocks, inter-frame prediction is performed for each division mode of each CTU block and reference frame information is obtained, and each division mode of each CTU block is intra-frame prediction And generate a prediction information corresponding to each division mode;

The intra-frame prediction coarse selection pipeline passes through the intra-frame prediction coarse selection module: performs intra-frame prediction on each PU block in each division mode and calculates the corresponding cost, and selects one or more costs relative to each PU block according to the cost. Intra prediction directions, and the selected intra prediction direction is used as the prediction information corresponding to the division mode.
The H.265 encoding device according to claim 24 or 25, wherein:

The inter-frame prediction coarse selection module includes: a coarse search module, a reference frame data loading module, a fine search module, and a fractional pixel search module;

The rough selection pipeline includes: rough search pipeline, reference frame data loading pipeline, fine search pipeline and fractional pixel search pipeline;

The coarse search pipeline stage passes through the coarse search module: select a frame from the reference array, select a reference frame from its original frame or reconstructed frame, perform down-sampling operations on the reference frame and the current CTU block, and perform down-sampling on the Find the pixel location with the least cost compared with the down-sampled CTU block in the reference frame, and calculate the coarse search vector of the pixel location relative to the current CTU block;

The reference frame data loading pipeline stage is through the reference frame data loading pipeline stage: the coarse search vector of the coarse search pipeline is obtained through the overall control module, and one or more predictions with the same function as the coarse search are obtained according to the motion vector around the CTU block Motion vector, load reference frame data according to the coarse search vector and one or more prediction vectors, and pass it to the fine search pipeline through the overall control module;

The fine search pipeline passes the fine search module: according to the coarse search vector, a fine search area is set in the reconstructed image of the reference frame for each PU block, and a corresponding PU block is generated in the fine search area A fine search vector with the smallest cost; and used to generate one or more predicted motion vectors with the same function as the coarse search vector based on the motion vector information around the current CTU block, and generate a fine search vector based on the predicted motion vector; and Send all the generated fine search vectors to the fractional pixel search module;

The fractional pixel search pipeline level passes through the fractional pixel search module: according to each received fine search vector, a corresponding fractional pixel search area is set in the reference frame for each PU block, and in the fractional pixel search area Generate a fractional pixel search vector with the smallest cost corresponding to the PU block.
The H.265 encoding device according to claim 26, wherein the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are the same pipeline, and the intra-frame prediction coarse selection module and the fractional pixel search Modules are executed in parallel at the same pipeline stage.
The H.265 encoding device according to claim 27, wherein:

The intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed in the intra-frame prediction coarse selection pipeline;

The rough selection pipeline of intra prediction includes: for each PU block in each division mode, the original pixels of the current frame are used to generate reference pixels, and all intra predictions are performed according to the rules of the H.265 protocol according to the reference pixels. Direction prediction is performed to obtain prediction results in each direction, and the distortion cost is calculated with the original pixels according to the prediction results in each direction, and one or more intra-frame prediction directions with lower cost are selected by sorting the cost from small to large.
The H.265 encoding device of claim 26, wherein the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are different pipeline stages, and the intra-frame prediction coarse selection module executes the fractional pixel search The pipeline stage after the module.
The H.265 encoding device according to claim 29, wherein:

The intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed in the intra-frame prediction coarse selection pipeline;

The reference pixel generation module is used to generate reference pixels using the reconstructed pixels of the current frame for each PU block in each division mode, and perform all intra prediction directions according to the rules of the H.265 protocol according to the reference pixels. The prediction results in each direction are obtained by prediction, the distortion cost is calculated with the original pixels according to the prediction results in each direction, and the cost is sorted from small to large to select one or more intra-frame prediction directions with a small cost.
The H.265 encoding device of claim 23, wherein the multiple modules further comprise a post-processing module, and the multiple pipeline steps further include a post-processing pipeline step,

The post-processing pipeline step passes through the post-processing module to accurately compare the partition mode with the lowest cost corresponding to each CTU block output by the module and generate a reconstructed frame corresponding to the current frame according to the corresponding reconstruction information.
The H.265 encoding device according to claim 23, wherein the multiple modules further comprise an entropy encoding module, and the multiple pipeline steps further include an entropy encoding pipeline step,

The entropy coding pipeline step uses the entropy coding module to accurately compare the partition mode with the lowest cost corresponding to each CTU block output by the module and generate a binary code stream conforming to the H.265 protocol specification according to the corresponding entropy coding information.
An H.265 encoding method, characterized in that the method is applied to an H.265 encoding device, the device includes multiple modules and multiple pipeline steps, and each pipeline step includes at least one pipeline stage for executing at least A module in which:

The multiple modules include a preprocessing module, a coarse selection module, an accurate comparison module, and an overall control module, and the overall control module is respectively connected to the preprocessing module, the rough selection module, and the precise comparison module;

The multiple pipeline steps include a pretreatment pipeline step, a rough selection pipeline step, and an accurate comparison pipeline step, the rough selection pipeline step is performed after the pretreatment pipeline step, and the precise comparison pipeline step is executed after the rough selection pipeline step;

The method includes the following steps:

The preprocessing pipeline step divides a current frame in an original video into multiple CTU blocks through the preprocessing module;

The rough selection pipeline process uses the rough selection module to divide each CTU block according to multiple division modes, and performs coarse selection of inter prediction and coarse selection of intra prediction for each division mode of each CTU block, and generates one and each Forecast information corresponding to the division mode;

The precise comparison pipeline step calculates and compares the prediction information corresponding to each partition mode of each CTU block through the precise comparison module, and selects the partition mode with the smallest cost for each CTU block and the partition mode corresponding to the partition mode. Encoding information, and according to the selected division mode and its corresponding encoding information, generating entropy encoding information for generating H.265 bitstream from the current frame and reconstruction information for generating reconstructed frames from the current frame,

The overall control module is used to control the storage and retrieval of original frame data and reference frame data, and to control the preprocessing module, the coarse selection module, and the precise comparison module to sequentially execute the corresponding pipeline steps.
The H.265 encoding method according to claim 33, wherein:

The coarse selection module includes: an inter prediction coarse selection module and an intra prediction coarse selection module; the coarse selection pipeline includes: an inter prediction coarse selection pipeline and an intra prediction coarse selection pipeline;

The method also includes:

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides each of them The CU block is divided into one or more corresponding PU blocks, and each division mode of each CTU block is inter-predicted and reference frame information is obtained, and each division mode of each CTU block is intra-predicted and generated A prediction information corresponding to each division mode;

Intra-frame prediction coarse selection pipeline through the intra-frame prediction coarse selection module: Perform intra-frame prediction for each PU block in each division mode and calculate the corresponding cost, and select one or more frames relative to the cost of each PU block according to the cost Intra prediction direction, and the selected intra prediction direction is used as the prediction information corresponding to the division mode.
The H.265 encoding method according to claim 33, wherein:

The coarse selection module further includes a coarse inter-frame selection module, and the precise comparison module further includes a coarse intra-frame selection module;

The rough selection pipeline includes a rough selection pipeline between frames, and the accurate comparison pipeline includes a rough selection pipeline within a frame;

The method includes:

The inter-frame prediction coarse selection pipeline uses the inter-frame prediction coarse selection module to divide each CTU block according to multiple division modes. Each division mode divides a CTU block into corresponding multiple CU blocks, and divides each of them The CU block is divided into one or more corresponding PU blocks, and each division mode of each CTU block is inter-predicted and reference frame information is obtained, and each division mode of each CTU block is intra-predicted and generated A prediction information corresponding to each division mode;

Intra-frame prediction coarse selection pipeline through the intra-frame prediction coarse selection module: Perform intra-frame prediction for each PU block in each division mode and calculate the corresponding cost, and select one or more frames relative to the cost of each PU block according to the cost Intra prediction direction, and the selected intra prediction direction is used as the prediction information corresponding to the division mode.
The H.265 encoding method according to claim 34 or 35, wherein:

The inter-frame prediction coarse selection module includes: a coarse search module, a reference frame data loading module, a fine search module, and a fractional pixel search module;

The rough selection pipeline includes: rough search pipeline, reference frame data loading pipeline, fine search pipeline, and fractional pixel search pipeline;

The method includes:

The coarse search pipeline stage passes through the coarse search module: select a frame from the reference array, select a reference frame in its original frame or reconstructed frame, perform down-sampling operations on the reference frame and the current CTU block, and perform the down-sampled reference frame Find the pixel location with the least cost compared with the down-sampled CTU block, and calculate the coarse search vector of the pixel location relative to the current CTU block;

The reference frame data loading pipeline stage The reference frame data loading pipeline stage: obtain the coarse search vector of the coarse search pipeline through the overall control module and obtain one or more predicted motion vectors with the same function as the coarse search according to the motion vectors around the CTU block , Load the reference frame data according to the coarse search vector and one or more prediction vectors, and pass it to the fine search pipeline through the overall control module;

The fine search pipeline stage passes through the fine search module: according to the coarse search vector, a fine search area is set in the reconstructed image of the reference frame for each PU block, and a cost corresponding to the PU block is generated in the fine search area The smallest fine search vector; and according to the motion vector information around the current CTU block, one or more predicted motion vectors with the same function as the coarse search vector are generated, and the fine search vector is generated based on the predicted motion vector; and all generated The refined search vector is sent to the fractional pixel search module;

The fractional pixel search pipeline level passes through the fractional pixel search module: according to each received fine search vector, a corresponding fractional pixel search area is set in the reference frame for each PU block, and in the fractional pixel search area Generate a fractional pixel search vector with the smallest cost corresponding to the PU block.
The H.265 encoding method of claim 36, wherein the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are the same pipeline, and the intra-frame prediction coarse selection module and the fractional pixel search Modules are executed in parallel at the same pipeline stage.
The H.265 encoding method according to claim 37, wherein:

The intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed in the intra-frame prediction coarse selection pipeline;

The method includes:

The rough selection pipeline of intra prediction includes: for each PU block in each division mode, the original pixels of the current frame are used to generate reference pixels, and all intra prediction directions are performed according to the rules of the H.265 protocol according to the reference pixels. The prediction results in each direction are obtained by prediction, the distortion cost is calculated with the original pixels according to the prediction results in each direction, and the cost is sorted from small to large to select one or more intra-frame prediction directions with a small cost.
The H.265 encoding method of claim 36, wherein the intra-frame prediction coarse selection pipeline and the fractional pixel search pipeline are different pipeline stages, and the intra-frame prediction coarse selection module executes the fractional pixel search The pipeline stage after the module.
The H.265 encoding method of claim 39, wherein:

The intra-frame prediction coarse selection module includes a reference pixel generation module, which is executed in the intra-frame prediction coarse selection pipeline;

The method includes:

The reference pixel generation module uses the reconstructed pixels of the current frame to generate reference pixels for each PU block in each division mode, and predicts all the intra-frame prediction directions according to the rules of the H.265 protocol according to the reference pixels to obtain each direction According to the prediction results of each direction, the distortion cost is calculated with the original pixels, and the cost is sorted from small to large to select one or more intra prediction directions with a small cost.
The H.265 encoding method according to claim 33, wherein the multiple modules further comprise a post-processing module, and the multiple pipeline steps further include a post-processing pipeline step,

The method includes:

The post-processing pipeline step uses the post-processing module to accurately compare the partition mode with the lowest cost corresponding to each CTU block output by the module and generate a reconstructed frame corresponding to the current frame according to the corresponding reconstruction information.
The H.265 encoding method according to claim 33, wherein the multiple modules further comprise an entropy encoding module, and the multiple pipeline steps further include an entropy encoding pipeline step;

The method includes: the entropy coding pipeline step, through the entropy coding module, accurately compares each CTU block output by the module with the least costly partition mode and generates a binary compliant with H.265 protocol specifications according to the corresponding entropy coding information Code stream.