阅读说明：本技术 一种视频编码的方法、系统和装置 (Method, system and device for video coding ) 是由 程志刚 贾春华 于 2021-07-30 设计创作，主要内容包括：本发明公开了一种视频编码的方法、系统和装置,属于视频编码的技术领域,所述方法包括帧间预测的方法：参考深度和参考代价体现了当前像素块与上游相邻像素块的相关性,通过该相关性,来确定当前像素块的最佳划分深度,降低四叉树划分复杂度；再结合当前像素块的类型、参考深度和参考代价,跳过一些不必要的帧间预测模式,从相应的列表中选择最优预测模式,降低因遍历全部模型率失真代价计算带来的计算量,从而降低视频编码过程中的运算量、并提高编码效率。(The invention discloses a method, a system and a device for video coding, belonging to the technical field of video coding, wherein the method comprises an interframe prediction method: the reference depth and the reference cost represent the correlation between the current pixel block and the upstream adjacent pixel block, and the optimal division depth of the current pixel block is determined through the correlation, so that the division complexity of the quadtree is reduced; and then, combining the type, the reference depth and the reference cost of the current pixel block, skipping some unnecessary inter-frame prediction modes, and selecting the optimal prediction mode from the corresponding list, so that the calculation amount caused by traversing all model rate distortion cost calculation is reduced, the calculation amount in the video coding process is reduced, and the coding efficiency is improved.)

1. A method of video coding, comprising a method of inter-prediction:

acquiring the type of a current pixel block, wherein the type comprises a static pixel block or a motion pixel block;

judging whether the type of the current pixel block is a static pixel block;

if yes, screening a prediction mode from a first list to predict, wherein the first list comprises: SKIP, Merge, inter2 Nx 2N, IBC and PLT;

if not, acquiring the depth and cost of an associated pixel block of the current pixel block, wherein the associated pixel block comprises a previous frame of co-located pixel block and a pixel block which is upstream of the current pixel block and is close to the current pixel block;

carrying out weighted sum on the depths of the associated pixel blocks to obtain a reference depth;

carrying out weighted sum on the costs of the associated pixel blocks to obtain a reference cost;

judging whether a second condition is met, wherein the reference cost is more than a second threshold value T2, the pixel block type is a motion pixel block, and the reference depth is less than a fourth threshold value T4;

if the second condition is satisfied, the optimal division depth of the current pixel block is 0 or 1.

2. The video encoding method of claim 1,

determining whether a third condition is satisfied, the third condition comprising: the reference cost is below a third threshold T3, the pixel block type is a motion pixel block, and the reference depth is below a fifth threshold T5;

if the third condition is satisfied, the optimal division depth of the current pixel block is 2 or 3.

3. The video coding method of claim 2,

if the second condition and the third condition are not met and the current pixel block is a motion pixel block;

judging whether the depth of the current pixel block is three;

if the depth is three, traversing an optimal prediction mode;

if the depth is less than three, dividing the current pixel block into the next depth;

judging whether the division depth of the current pixel block is the optimal depth;

if so, acquiring an optimal prediction model;

if the depth is not the optimal depth, the next depth division is performed.

4. The video coding method of claim 1, wherein the method for obtaining the type of the current pixel block comprises:

obtaining the pixel value difference D between the pixel point of the current pixel block and the same-position pixel point of the previous frame_k：

D_k＝|f_k(x,y)-f_k-1(x,y)|

Wherein f is_k(x, y) is expressed as a pixel value of a pixel point, f_k-1(x, y) representing the same-position pixel point in the previous frame;

if the pixel value difference is zero, the pixel point is a background point;

judging whether the proportion of the background points is greater than a first threshold value or not;

if the current pixel block is larger than the first threshold value, the current pixel block is a static pixel block;

if the pixel value is less than or equal to the first threshold value, the current pixel block is a motion pixel block.

5. The video coding method of claim 1, wherein the reference depth is calculated as:

lp＝λ₁×l_i,j,f-1+λ₂×l_i,j-1,f+λ₃×l_i-1,j,f+λ₄×l_i-1,j-1,f+λ₅×l_i-1,j+1,f

where lp is expressed as the reference depth, λ₁、λ₂、λ₃、λ₄And λ₅Is a weight coefficient, l_i,j,fIs shown asWhen the depth of the row pixel blocks, f is expressed as a frame, j is expressed as a column, and i is expressed as a row;

the calculation formula of the reference cost is as follows:

RD_ave＝λ₁×RD_i,j,f-1+λ₂×RD_i,j-1,f+λ₃×RD_i-1,j,f+λ₄×RD_i-1,j-1,f+λ₅×RD_i-1,j+1,f

wherein RD_aveExpressed as a reference cost, RD, of the current block of pixels_i，j，f-1Representing the cost of the co-located pixel block of the previous frame.

6. The video coding method of claim 1, further comprising a method of motion pixel block prediction model selection:

if the current partition depth is 0 or 1, selecting the best prediction model from a second list, wherein the second list comprises: the best prediction models of SKIP, Merge, inter2N multiplied by 2N, IBC, PLT and the associated pixel blocks;

if the current division depth is 2 or 3, acquiring the number of corner points of the current pixel block;

if the number of the angular points is more than 0, traversing an optimal prediction mode;

and if the number of the corner points is 0, skipping the IBC and the PLT, and traversing the optimal prediction model.

7. The video coding method of claim 6, wherein the corner point detection method comprises:

acquiring the brightness of a pixel core point of a current pixel block;

and if the brightness difference between the pixel point of the current pixel block and the pixel core point is greater than or equal to the sixth threshold, the pixel point is an angular point.

8. The video coding method of claim 1, wherein inter prediction obtains a prediction residual between a current pixel block and an adjacent pixel block;

transforming the prediction residual error to obtain a transformation coefficient;

and after the transformation coefficient is quantized, entropy coding is carried out.

9. A system for implementing the video coding method according to any one of claims 1 to 8, comprising a pixel block classification module, an associated pixel block analysis module, a filtering module and a depth analysis module,

the pixel block classification module is used for acquiring the type of the current pixel block;

the associated pixel block analysis module is used for acquiring the depth and cost of an associated pixel block of a current pixel block, carrying out weighted sum on the depth to acquire reference depth, and carrying out weighted sum on the cost to acquire reference cost;

the screening module is used for: when the category of the current pixel block is a static pixel block, screening a prediction mode from the first list for prediction;

the depth analysis module is used for: when the second condition is satisfied, the optimal division depth of the current pixel block is 0 or 1.

10. An apparatus for video encoding comprising a processor, a memory, and a program stored in the memory,

the program comprising instructions for implementing a video encoding method according to any one of claims 1 to 8;

the processor is configured to execute the program.

Technical Field

The present invention relates to the field of video coding technologies, and in particular, to a method, a system, and an apparatus for video coding.

Background

The screen video is displayed by electronic products such as computers, televisions, mobile phones, tablet computers and the like through a screen. Screen videos have been widely used in the office and surveillance fields, such as remote training videos, desktop sharing, desktop collaboration, screen recording of surveillance videos, and the like.

The HEVC/h.265 coding structure uses a quadtree structure and larger Coding Units (CUs), as shown in fig. 3, which results in significantly increased computational complexity of the encoder and difficulty in meeting the real-time requirement of the encoding time. To improve compression efficiency, h.265 encoders have more flexible pixel block (CU) partitioning. Each frame image is first sequentially divided into LCUs of 64 × 64 size, and each CU may be recursively divided into CUs of 4 depths (64 × 64, 32 × 32, 16 × 16, and 8 × 8) starting from the LCUs and having coding depths of 0 to 3, to construct a quad-tree coding structure. I.e., a depth of 0, a CU corresponds to an LCU. In prediction, each pixel block has a corresponding prediction mode, and residual estimation is performed, wherein the residual can be motion estimation and motion compensation.

The prediction mode of h.265 increases the complexity of the whole coding process, and in the CU division process, in order to determine the quadtree structure of a CU, a complete traversal of the depth l from 0 to 3 needs to be performed, and 4 is calculated in total⁰+4¹+4²+4³The calculation process is complicated because of the rate distortion cost of 85 times. Meanwhile, the pixel block needs to traverse all the prediction modes, and selects the prediction mode with the minimum coding cost as the optimal prediction mode. Taking a 64 × 64 LCU as an example, the prediction mode needs to traverse 1662-2216 times, and the calculation complexity of the prediction mode selection in the whole video encoding process is very high, obviously, the calculation complexity of the encoding end is very high due to the traversal calculation process, and the encoding time consumed by video compression is long, so that the increasing real-time video compression requirement cannot be met.

Therefore, a large amount of operation complexity is introduced in the prediction process of the HEVC/h.265, and how to effectively reduce the operation amount of video coding becomes a problem to be solved urgently at present.

Disclosure of Invention

In view of the above technical problems in the prior art, the present invention provides a method, a system and a device for video encoding, which reduce the amount of computation in the video encoding process and improve the encoding efficiency.

The invention discloses a video coding method, which comprises an interframe prediction method:

acquiring the type of a current pixel block, wherein the type comprises a static pixel block or a motion pixel block;

judging whether the type of the current pixel block is a static pixel block;

if yes, screening a prediction mode from a first list to predict, wherein the first list comprises: SKIP, Merge, inter2 Nx 2N, IBC and PLT;

if not, acquiring the depth and cost of an associated pixel block of the current pixel block, wherein the associated pixel block comprises a previous frame of co-located pixel block and a pixel block which is upstream of the current pixel block and is close to the current pixel block;

carrying out weighted sum on the depths of the associated pixel blocks to obtain a reference depth;

carrying out weighted sum on the costs of the associated pixel blocks to obtain a reference cost;

judging whether a second condition is met, wherein the reference cost is more than a second threshold value T2, the pixel block type is a motion pixel block, and the reference depth is less than a fourth threshold value T4;

if the second condition is satisfied, the optimal division depth of the current pixel block is 0 or 1.

Preferably, it is determined whether a third condition is satisfied, the third condition including: the reference cost is above the third threshold T3, the pixel block type is a motion pixel block, and the reference depth is below the fifth threshold T5;

if the third condition is satisfied, the optimal division depth of the current pixel block is 2 or 3.

Preferably, if the second condition and the third condition are not satisfied and the current pixel block is a motion pixel block;

judging whether the depth of the current pixel block is three;

if the depth is three, traversing an optimal prediction mode;

if the depth is less than three, dividing the current pixel block into the next depth;

judging whether the division depth of the current pixel block is the optimal depth;

if so, acquiring an optimal prediction model;

if the depth is not the optimal depth, the next depth division is performed.

Preferably, the method for obtaining the type of the current pixel block comprises:

obtaining the pixel value difference D between the pixel point of the current pixel block and the same-position pixel point of the previous frame_k：

D_k＝|f_k(x,y)-f_k-1(x,y)|

Wherein f is_k(x, y) is expressed as a pixel value of a pixel point, f_k-1(x, y) representing the same-position pixel point in the previous frame;

if the pixel value difference is zero, the pixel point is a background point;

judging whether the proportion of the background points is greater than a first threshold value or not;

if the current pixel block is larger than the first threshold value, the current pixel block is a static pixel block;

if the pixel value is less than or equal to the first threshold value, the current pixel block is a motion pixel block.

Preferably, the reference depth is calculated by the formula:

lp＝λ₁×l_i,j,f-1+λ₂×l_i,j-1,f+λ₃×l_i-1,j,f+λ₄×l_i-1,j-1,f+λ₅×l_i-1,j+1,f

where lp is expressed as the reference depth, λ₁、λ₂、λ₃、λ₄And λ₅Is a weight coefficient, l_i,j,fExpressed as the depth of the pixel blocks in the row, f as the frame, j as the column, and i as the row.

Preferably, the reference cost is calculated by the formula:

RD_ave＝λ₁×RD_i,j,f-1+λ₂×RD_i,j-1,f+λ₃×RD_i-1,j,f+λ₄×RD_i-1,j-1,f+λ₅×RD_i-1,j+1,f

wherein RD_aveExpressed as a reference cost, RD, of the current block of pixels_i，j，f-1Representing the cost of the co-located pixel block of the previous frame.

Preferably, the method further comprises the following steps:

if the current partition depth is 0 or 1, selecting the best prediction model from a second list, wherein the second list comprises: the best prediction models of SKIP, Merge, inter2N multiplied by 2N, IBC, PLT and the associated pixel blocks;

if the current division depth is 2 or 3, acquiring the number of corner points of the current pixel block;

if the number of the angular points is more than 0, traversing an optimal prediction mode;

and if the number of the corner points is 0, skipping the IBC and the PLT, and traversing the optimal prediction model.

Preferably, the method for detecting the corner point includes:

acquiring the brightness of a pixel core point of a current pixel block;

and if the brightness difference between the pixel point of the current pixel block and the pixel core point is greater than or equal to the sixth threshold, the pixel point is an angular point.

Preferably, the inter-frame prediction obtains a prediction residual between the current pixel block and the adjacent pixel block; transforming the prediction residual error to obtain a transformation coefficient; and after the transformation coefficient is quantized, entropy coding is carried out.

The invention also provides a system for realizing the video coding method, which comprises a pixel block classification module, an associated pixel block analysis module, a screening module and a depth analysis module,

the pixel block classification module is used for acquiring the type of the current pixel block;

the associated pixel block analysis module is used for acquiring the depth and cost of an associated pixel block of a current pixel block, carrying out weighted sum on the depth to acquire reference depth, and carrying out weighted sum on the cost to acquire reference cost;

the screening module is used for: when the category of the current pixel block is a static pixel block, screening a prediction mode from the first list for prediction;

the depth analysis module is used for: when the second condition is satisfied, the optimal division depth of the current pixel block is 0 or 1.

Preferably, the system of the present invention further comprises a partitioning module,

the dividing module is configured to: and when the current division depth is smaller than the optimal division depth, performing next-level depth division on the current pixel block.

The invention also provides a video coding device, which comprises a processor, a memory and a program stored in the memory, wherein the program comprises instructions for realizing the video coding method; the processor is configured to execute the program.

Compared with the prior art, the invention has the beneficial effects that: the reference depth and the reference cost represent the correlation between the current pixel block and the upstream adjacent pixel block, and the optimal division depth of the current pixel block is determined through the correlation, so that the division complexity of the quadtree is reduced; and then, combining the type, the reference depth and the reference cost of the current pixel block, skipping some unnecessary inter-frame prediction modes, and selecting the optimal prediction mode from the corresponding list, so that the calculation amount caused by traversing all model rate distortion cost calculation is reduced, the calculation amount in the video coding process is reduced, and the coding efficiency is improved.

Drawings

FIG. 1 is a flow chart of a video encoding method of the present invention;

FIG. 2 is a schematic diagram of the positional relationship of associated pixel blocks;

FIG. 3 is a schematic diagram of a pixel core point;

FIG. 4 is a logical block diagram of the system of the present invention;

fig. 5 is a flow chart of a method of prediction model selection for a block of motion pixels.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.

The invention is described in further detail below with reference to the attached drawing figures:

a method of video coding, as shown in fig. 1, includes a method of inter-prediction:

step 101: obtaining the type of the current pixel block, wherein the type comprises a static pixel block or a motion pixel block. Coding units in the pixel block HEVC include Coding Tree Units (CTUs) and Coding Units (CUs). A still pixel block refers to a pixel having a small change in pixel value, and a moving pixel block refers to a pixel having a large change in pixel value.

Step 102: and judging whether the type of the current pixel block is a static pixel block.

If yes, go to step 103: the current depth is the best depth, and a prediction mode is screened from a first list for prediction, wherein the first list comprises: SKIP, Merge, inter2N × 2N, IBC, and PLT.

If not, go to step 104: and acquiring the depth and cost of an associated pixel block of the current pixel block, wherein the associated pixel block comprises a previous frame of co-located pixel block and a pixel block which is upstream of the current pixel block and is close to the current pixel block. As shown in FIG. 2, the associated pixel block is typically located upstream of the current pixel block, and includes the previous frame co-located pixel block, the left pixel block, the upper right pixel block, and the depth and cost thereof are typically calculated prior to the current pixel block.

Step 105: and carrying out weighted sum on the depths of the associated pixel blocks to obtain a reference depth lp.

Step 106: weighting and summing the associated pixel block cost to obtain a reference cost RD_ave。

Step 107: and judging whether the second condition is met, wherein the reference cost is more than or equal to a second threshold value T2, the pixel block type is a motion pixel block, and the reference depth is less than or equal to a fourth threshold value T4.

If the second condition is satisfied, execute step 108: the current pixel block has the best depth of 0 or 1, and the division with the depths of 2 and 3 is skipped. When the division of depths 2 and 3 is skipped, that is, when the depth is 1, the division of the next depth is not performed.

The reference depth and the reference cost represent the correlation between the current pixel block and the upstream adjacent pixel block, and the optimal division depth of the current pixel block is determined through the correlation, so that the division complexity of the quadtree is reduced; and then, combining the type, the reference depth and the reference cost of the current pixel block, skipping some unnecessary inter-frame prediction modes, and selecting the optimal prediction mode from the corresponding list, so that the calculation amount caused by traversing all model rate distortion cost calculation is reduced, the calculation amount in the video coding process is reduced, and the coding efficiency is improved.

If the second condition is not satisfied, go to step 110: and judging whether the third condition is met, wherein the reference cost is below a third threshold value T3, the pixel block type is a motion pixel block, and the reference depth is below a fifth threshold value T5.

If the third condition is satisfied, execute step 111: the optimal depth of the current block of pixels is 2 or 3.

Step 120: and if the second condition and the third condition are not met and the current pixel block is the motion pixel block.

Step 121: and judging whether the depth of the current pixel block is three.

If the depth is three, go to step 122: and traversing the optimal prediction mode.

If the depth is less than three, go to step 123: and dividing the current pixel block into the next depth.

According to the obtained optimal depth, guidance can be provided for the division of the pixel block: judging whether the division depth of the current pixel block is the optimal depth; if so, acquiring an optimal prediction model; and if the depth is not the optimal depth, the next depth is divided, and the screening of the optimal prediction model and the cost calculation are skipped.

In step 101, obtaining the type of the current pixel block includes:

obtaining the pixel value difference D between the pixel point of the current pixel block and the same-position pixel point of the previous frame_k：

D_k＝|f_k(x,y)-f_k-1(x,y)|

Wherein f is_k(x, y) is expressed as pixel value, f_k-1(x, y) is expressed as the pixel value of the same-position pixel point in the previous frame;

if the pixel value is different from D_kZero, the pixel point is a background point;

judging whether the proportion of the background points is greater than a first threshold value T1;

if the current pixel block is larger than the first threshold value T1, the current pixel block is a static pixel block;

if the pixel value is less than or equal to the first threshold value T1, the current pixel block is a motion pixel block.

Namely, the type of the current pixel block is judged by the ratio of the background points.

In step 103, the calculation formula of the reference depth is:

lp＝λ₁×l_i,j,f-1+λ₂×l_i,j-1,f+λ₃×l_i-1,j,f+λ₄×l_i-1,j-1,f+λ₅×l_i-1,j+1,f

where lp is expressed as the reference depth, λ₁、λ₂、λ₃、λ₄And λ₅Is a weight coefficient, l_i，j，fExpressed as the depth of the pixel block in the line, f is the frame, j is the column, i is the line, l is the line_i，j，f-1Expressed as the depth of the co-located pixel block of the previous frame,/_i，j-1，fExpressed as the depth, l, of the left pixel block_i-1，j，fExpressed as the depth, l, of the upper pixel block_{i-1，j-1，f}Expressed as the depth, l, of the upper left pixel block_{i-1，j+1，f}Indicated as the depth of the upper right pixel block, fig. 2 shows the positional relationship of the associated pixel blocks.

The calculation formula of the reference cost is as follows:

RD_ave＝λ₁×RD_i,j,f-1+λ₂×RD_i,j-1,f+λ₃×RD_i-1,j,f+λ₄×RD_i-1,j-1,f+λ₅×RD_i-1,j+1,fwherein RD_aveExpressed as a reference cost, RD, of the current block of pixels_i，j，f-1Expressed as the cost, RD, of the co-located pixel block of the previous frame_i，j-1，fExpressed as the cost of the left pixel block, RD_i-1，j，fExpressed as the cost, RD, of the upper pixel block_{i-1，j-1，f}Expressed as cost, RD, of the upper left pixel block_{i-1，j+1，f}Expressed as the cost of the upper right pixel block. In a toolIn the bulk embodiment, λ₁Values of 0.3, lambda₂Values of 0.2, lambda₃Value, λ₄Values of 0.15 and λ₅The value of 0.15, that is, the distance between the upper left pixel block and the upper right pixel block and the current pixel block is relatively long, and the weight is relatively low, but the value of the weight coefficient is not limited thereto. Wherein λ is₁、λ₂、λ₃、λ₄、λ₅The sum of which is 1 and the reference depth and the reference cost are weighted averages of upstream neighboring pixel blocks. Wherein, the associated pixel block is an upstream pixel block, the cost and the depth are already calculated, and therefore, corresponding data can be read.

In a specific embodiment, T1 takes 12%, T2 takes 12%, T3 takes 2%, T4 takes 2.5, and T5 takes 0.8, but is not limited thereto.

The prediction modes include: SKIP mode, Merge mode, rectangular motion split mode (Square), symmetric motion Split Mode (SMP), asymmetric motion split mode (AMP), and inter mode (intra mode), the rectangular motion split mode including inter2N × 2N and inter N × N; the symmetrical motion segmentation mode comprises inter2 NxN and inter Nx2N; the asymmetric motion segmentation mode comprises inter2 NxnU, inter2 NxnD, inter nL x 2N and inter nR x 2N; the inter mode includes intra 2N × 2N and intra N × N; according to the inter-frame coding method for the IBC and PLT screen contents, an intra-frame block copy mode (IBC) can perform matching block search in a reconstruction area of a current frame, so that the coding performance of a video is effectively improved, and a video block is mapped into a color table and an index map by a palette mode (PLT).

As shown in fig. 5, the method of the present invention further comprises a method of prediction model selection of a motion pixel block:

step 201: and acquiring the division depth of the motion pixel block.

Step 202: and judging whether the current division depth is 0 or 1.

If yes, go to step 203: selecting a best prediction model from a second list, the second list comprising: SKIP, Merge, inter2N × 2N, IBC, PLT and the best prediction model for the associated pixel block. The associated pixel block is an upstream pixel block, which is usually already coded, and its optimal prediction model is also obtained.

If not, that is, the current partition depth is 2 or 3, execute step 204: the number of corner points of the current pixel block is obtained.

Step 205: and judging whether the number of corner points is more than 0.

If yes, i.e. the number of corner points is greater than 0, go to step 206: and traversing the optimal prediction mode.

If not, that is, the number of corner points is 0, execute step 207: IBC and PLT are skipped and the optimal prediction model is traversed. For a motion pixel block (CU) where no corner is detected, which indicates that the motion pixel block is of a conventional natural video type, IBC and PLT modes can be skipped to enter the selection of a conventional intra or inter prediction model. Namely, IBC and PLT modes are eliminated, and the optimal prediction model is screened from the SKIP mode, the Merge mode, the rectangular motion segmentation mode, the symmetric motion segmentation mode and the asymmetric motion segmentation mode.

The method for detecting the corner comprises the following steps:

acquiring the brightness of a pixel core point of a current pixel block;

and if the brightness difference between the pixel point of the current pixel block and the pixel core point is greater than or equal to the sixth threshold, the pixel point is an angular point.

Corner detection can be expressed by the following equation:

wherein C is expressed as a comparison function, C ═ 1 denotes that a pixel point is an angular point, r0 denotes a pixel core point of a pixel block, r denotes other pixel points, I denotes the luminance of the pixel point, and T6 denotes a sixth threshold value.

Fig. 3 shows a specific pixel block, where the pixel point numbered 19 is the pixel core point. The corner detection is obtained by a method based on a minimum homological region (USAN, united value Segment acquisition Nucleus).

The present invention also provides a system for implementing the above video coding method, as shown in fig. 4, comprising a pixel block classification module 1, an associated pixel block analysis module 2, a screening module 3 and a depth analysis module 4,

the pixel block classification module 1 is used for acquiring the type of the current pixel block;

the associated pixel block analysis module 2 is configured to obtain a depth and a cost of an associated pixel block of a current pixel block, perform a weighted sum on the depth to obtain a reference depth, and perform a weighted sum on the cost to obtain a reference cost;

the screening module 3 is used for: when the category of the current pixel block is a static pixel block, screening a prediction mode from the first list for prediction;

the depth analysis module 4 is configured to: when the second condition is satisfied, the optimal division depth of the current pixel block is 0 or 1.

The system of the present invention may further include a partitioning module 5, where the partitioning module 5 is configured to perform the next-level depth partitioning on the current pixel block when the current partitioning depth is smaller than the optimal partitioning depth.

The screening module 3 is further configured to screen the prediction model from the second list, or traverse the optimal prediction model.

The system of the invention also comprises a prediction module 11, a transformation module 12 and an encoding module 13;

the prediction module 11 is configured to obtain a prediction residual between a current pixel block and an adjacent pixel block based on a prediction model, and obtain the prediction residual by inter-frame predicting the current pixel block;

the transformation module 12 is configured to transform the prediction residual to obtain a transformation coefficient;

the coding module 13 is configured to perform entropy coding after quantizing the transform coefficient, so as to complete coding of the current pixel block.

The invention also provides a video coding device, which comprises a processor, a memory and a program stored in the memory, wherein the program comprises instructions for realizing the video coding method; the processor is configured to execute the program.

Examples

The environment adopted in this embodiment is Windows server 2008R2, the processor is Intel E5-2620CPU @2.1GHz, the single processor is 8 cores, and the RAM is 32 GB.

The video coding test is carried out by comparing various videos, and according to the video rapid coding method, video compression software H is designed_newIn contrast to HM12.0 in h.265, the algorithm performance is determined by PSNR (peak signal-to-noise ratio), a parameter for objectively evaluating video coding quality, and a time variation ratio Δ T (Δ T ═ T_HM-T_new)/T_HMX 100%) of two parameters, where T_HMTime encoded for HM 12.0; t is_newIs H_newThe time of the encoding.

The values of the respective thresholds are as follows: t1-12%, T2-12%, T3-2%, T4-2.5, and T5-0.8. The test results are shown in the following table:

the table shows that the PSNR is reduced by only 0.08dB on average, the video coding quality is almost unchanged, the video coding time is greatly saved, and the average value Δ T is 30.5%, which indicates that the video coding method of the present invention can reduce the video coding time consumption and improve the coding efficiency while effectively ensuring the coding quality by coding the video under the conditions of increasing the CU partition rate and selecting the prediction mode.

In this embodiment, various monitoring videos are played through the computer terminal, and secondary encoding is performed through video encoding. Video generally does not contain noise, many of which are computer-generated images or video, discrete in hue and less in number of colors; the lines of the screen image are finer and smoother, and the edges are sharper; meanwhile, most screen images have more uniform flat areas and more repeated patterns and the same blocks.

The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.