CN104539954A - Cascading method for speeding up high efficiency video coding (HEVC) - Google Patents

Abstract

The invention discloses a cascading method for speeding up high efficiency video coding (HEVC). The cascading method comprises the steps of precoding a current frame and a HEVC reestablished frame by H.264/AVC, and picking out a video macroblock partition result from a precoding result; then mapping the macroblock partition result of the H.264 to a possible mode set of the HEVC through a mode mapping; and finally, searching the best mode of the HEVC in the set by rate distortion optimization. Compared with prior art, the innovative cascading method for speeding up HEVC can greatly simplify the mode searching process of HEVC, and has good application prospect.

Description

A cascading method to speed up HEVC encoding

technical field

The invention relates to a method in the field of HEVC mode rapid selection, in particular to a method for accelerating HEVC encoding by cascading H.264 and HEVC encoders.

Background technique

HEVC is the abbreviation of High Efficiency Video Coding Standard. It is a new generation of video coding standard after H.264/AVC. Collaborative Team on Video Coding) and released in early 2013. HEVC adopts a block-based hybrid coding framework, which mainly includes technologies such as intra prediction, inter prediction, transformation and quantization, entropy coding, inverse quantization and inverse transformation, and loop filtering. Compared with the traditional coding framework, HEVC adopts new technologies such as flexible quadtree coding block division, large block prediction and transformation, multiple types of prediction blocks and transformation blocks, adaptive motion parameter coding, and adaptive loop filtering. . These new coding techniques effectively improve the coding performance of HEVC, but also introduce a huge amount of calculation, which is not suitable for practical applications.

In the mode selection process of HEVC, the standard model method is to exhaustively enumerate all possible combinations of coding units, prediction units and transformation units, and calculate the least rate-distortion to select the optimal coding mode. This method consumes a lot of calculations on the last modes that prove to be not optimal, so in order to promote the development of HEVC and the key to practical applications, it is very necessary to improve its mode selection method while maintaining the performance of HEVC video coding of. In general, improvements in mode selection are achieved by exploiting the temporal and spatial redundancies that exist among coding units, prediction units, and transform units. The existence of this redundancy is hidden and needs to be continuously explored, and the optimization and improvement of the mode selection method has important practical significance for the actual application of HEVC.

Contents of the invention

Aiming at the deficiencies of the prior art, the present invention provides a cascading method for accelerating HEVC encoding speed, adopts a brand-new encoder framework to solve the problem of fast selection of HEVC mode, and converts the H.264/AVC mode through the method of mode mapping Combining the macroblock partition results with the HEVC possible CU (Coding Unit) mode selection set provides a new way of thinking for HEVC's fast mode selection problem.

In order to achieve the above object, the present invention adopts the following technical solutions: the present invention designs a new frame mode of the encoder to accelerate the encoding speed of HEVC, that is: first use H.264/AVC to reconstruct the frame of the current frame and HEVC Pre-encoding is performed, and the macroblock division result of the video is extracted from the pre-encoding result; then the macroblock division result of H.264 is mapped to the possible mode set of HEVC through mode mapping; finally, HEVC is searched using rate-distortion optimization in this set best mode.

The cascading method of accelerating HEVC coding speed described in the present invention, the method comprises the following steps:

Step 1, precoding the current frame to obtain the optimal macroblock division;

First, the H.264 encoder is used to pre-encode the input frame to obtain the optimal macroblock division result of the frame. In the pre-encoding process, the framework of the encoder adopts the H.264/AVC hybrid coding framework, and the input parameters are the current frame and the reconstructed frame of HEVC, and the reconstructed frame of HEVC will be used as the reference frame of the current frame for motion estimation and macroblock division.

Step 2. Map the optimal macroblock division of H.264/AVC to the HEVC mode set

The result of optimal macroblock division obtained by H.264/AVC is mapped to the optimal possible mode set of HEVC by a mapping method. For CUs of different sizes, different mode mapping strategies are adopted to select the most appropriate CU mode set;

B. 64×64 layers

For 64×64 CUs, only SKIP and 2N×2N prediction modes are used, because experiments have shown that only using these two modes has little impact on the performance of the encoder, but it greatly reduces the complexity of the encoder.

C. 32x32 layers

For the 32×32 layer, the CU is divided into five different types according to the types of the four MBs in the CU. For each type of CU, different mapping relationships are designed to ensure the optimal set of possible mapping modes. The modes of the five CUs and their corresponding mappings are as follows:

a) When the current CU contains 3 or 4 MBs (macroblocks) whose mode is Inter16×16, only the Inter2N×2N mode of the current CU is detected.

b) When the modes of the two left MBs in the current CU are both Inter16×16, only the Inter2N×2N and Inter2N×N modes of the current CU are detected.

c) When the modes of the two right MBs in the current CU are both Inter16×16, only the Inter2N×2N and InterN×2N modes of the current CU are detected.

d) When the modes of the four MBs in the current CU are all Intra16×16, only the Intra2N×2N mode of the current CU is detected.

e) For other cases, all modes will be checked to determine the most suitable prediction mode for the current CU.

D. 16×16 and 8×8 layers

For 16×16 and 8×8 CUs, the present invention adopts different mapping relationships to obtain the most suitable PU set. Among them, for 16×16 and 8×8 CUs, the following mapping relationships are adopted respectively:

a) For 16×16 CU, H.264/AVC SKIP mode is mapped to HEVC SKIP/Merge mode; 16×16 mode is mapped to 2N×2N and SKIP/Merge mode; H.264/AVC 16×8 and 8×16 respectively correspond to HEVC’s SKIP/Merge, 2N×2N and corresponding left/right/upper and lower rectangle and asymmetric division modes; for 8×8, HEVC will detect all modes except Intra and N×N mode, for H Intra mode in .264/AVC, HEVC will detect all modes.

b) For 8×8 CU, H.264/AVC SKIP mode is mapped to HEVC SKIP/Merge mode; 16×16 mode is mapped to 2N×2N and SKIP/Merge mode; H.264/AVC 16×8 and 8×16 respectively correspond to HEVC’s SKIP/Merge, 2N×2N and corresponding left/right/upper and lower rectangle and asymmetric division modes; for 8×8, HEVC will detect all modes except Intra mode, for H.264/AVC Intra mode in HEVC will detect all modes.

Step 3. Use the HEVC encoder to traverse the PU mode sets that map different sizes of CUs, and use rate-distortion optimization to obtain the best CU and PU modes.

Compared with the prior art, the present invention has the following beneficial effects:

The present invention uses H.264/AVC to perform precoding to obtain the macroblock division result of the current frame; the division result of H.264/AVC is mapped to the CU mode set of HEVC through the method of mode mapping; finally, HEVC rate-distortion optimization is used to optimize the The optimal CU and PU (prediction unit) modes are selected in the mode set. The invention narrows down the HEVC mode set that needs to be searched through the mode mapping method, accelerates the encoding speed of the HEVC, utilizes a mature H.264/AVC encoder, and has a good application prospect.

Description of drawings

Other characteristics, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:

Figure 1 is a schematic diagram of four different CU (32×32) types;

Figure 2 is a block diagram of an encoder using cascaded H.264/AVC for accelerating HEVC encoding speed;

Fig. 3 is an illustration during the implementation of the present invention.

Detailed ways

The present invention will be described in detail below in conjunction with specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

Here, the application of the present invention in accelerating the encoding speed of HEVC is described. Specifically, the framework of the concatenated encoder constructed by the present invention and the mode mapping between H.264/AVC and HEVC are applied to the encoding process of HEVC. Specifically The block diagram is shown in Figure 1: firstly, the steps of precoding using H.264/AVC are introduced, and then the mode mapping method of different CU sizes will be introduced in detail on this basis, and finally, the selection method of the optimal mode of HEVC will be introduced.

Step 1, precoding the current frame to obtain the optimal macroblock division;

The encoding frameworks of HEVC and H.264/AVC are both hybrid encoding frameworks, and H.264/AVC has made great progress after years of development, and has done a lot of excellent work in codec optimization. Therefore, in the present invention, we first use the H.264 encoder to pre-encode the input frame to obtain the optimal macroblock division result of the frame. In the pre-encoding process, the encoder framework adopts the H.264/AVC hybrid encoding framework, and the input parameters are the current frame and the reconstructed frame of HEVC, and the reconstructed frame of HEVC will be used as the reference frame of the current frame.

Step 2. The optimal macroblock division of H.264/AVC is mapped to the HEVC mode set;

In order to improve coding efficiency, HEVC adopts a coding tree unit (CTU) based on an adaptive quadtree structure to replace the macroblock-based coding unit of H.264. In HEVC, the size of the CTU is 64×64, and it can be adaptively and iteratively quartered until the SCU (smallest coding unit), and the size of the SCU is 8×8. In H.264/AVC, the size of a macroblock (MB) can be 16×16, 8×8, 4×4. In H.264/AVC, inter-frame predictive coding only supports 7 macroblock partition modes: 16×16, 16×8, 8×16, 8×8, 8×4, 4×8 and 4×4, while HEVC not only supports these symmetric partition modes, called 2N×2N, 2N×N, and N×2N, but also supports asymmetric partition modes, called: 2N×nU, 2N×nD, nL×2N, and nR×2N . Compared with H.264 coding units with a fixed macroblock size, HEVC is more flexible in dividing coded images, thus providing better video image quality. This difference also directly leads to the high cost of transcoding from H.264 to HEVC. the complexity.

In the present invention, the result of the optimal macroblock division obtained by H.264/AVC is mapped to the optimal mode set possible for HEVC through the mapping method. For CUs of different layers, different mapping methods are selected to obtain the optimal mode set. the result of;

A. 64×64 layers

For 64×64 CUs, we only use SKIP and 2N×2N prediction modes, because experiments have shown that only using these two modes has little impact on the performance of the encoder, but it greatly reduces the complexity of the encoder.

B. 32x32 layers

For the 32×32 layer, we divide the CU into five different types according to the types of the four MBs in the CU. For each type of CU, we will have different mapping relationships to ensure the optimal effect. The modes of the five CUs and their corresponding mappings are as follows:

a) When the current CU contains 3 or 4 MBs whose mode is Inter16×16, we only detect the Inter2N×2N mode of the current CU.

b) When the modes of the two left MBs in the current CU are both Inter16×16, we only detect the Inter2N×2N and Inter2N×N modes of the current CU.

c) When the modes of the two right MBs in the current CU are both Inter16×16, we only detect the Inter2N×2N and InterN×2N modes of the current CU.

d) When the modes of the four MBs in the current CU are all Intra16×16, we only detect the Intra2N×2N mode of the current CU.

e) For other cases, all modes will be checked to determine the most suitable prediction mode for the current CU.

C. 16×16 and 8×8 layers

For 16×16 and 8×8 CUs, the present invention adopts different mapping relationships to obtain the most suitable PU set. The specific mapping modes are shown in Table 1 and Table 2:

Table 1: Mapping relation table between MB and CU (16×16)

Table 2: Mapping relation table between MB and CU (8×8)

Step 3. After obtaining the best possible mode set of HEVC, use the HEVC encoder to traverse the PU mode set that maps different sizes of CUs, calculate the rate-distortion cost of different modes, and use the rate-distortion optimization to determine the best CU and PU model.

Table 3 below is the running result of the present invention in X265. X265 is currently one of the open source encoders with outstanding HEVC performance. X265 fully supports the HEVC standard and has been deeply optimized in many aspects. At the same time, x264 is selected as the H.264/AVC encoder to pre-encode the current frame. X264 is the current open source H.264/AVC encoder with the best performance.

The test sequence set is processed by using the mode mapping method of the cascaded encoder of the present invention, and the experimental results show that the method of the present invention can map out the optimal possible mode set, which also has a good improvement in efficiency and accelerates the encoding speed of HEVC . Table 3 shows the performance and time variation of the method of the present invention for different test sequences.

Table 3: Experimental results based on X265 and X264

When the method of the present invention is used for HEVC encoding, the macroblock division result of H.264/AVC can be effectively used, and the macroblock division result is mapped to a possible mode set that can be used for HEVC CU mode selection, reducing the CU and PU mode to achieve the purpose of accelerating HEVC encoding.

The above descriptions are only preferred implementations of the present invention, and the protection scope of the present invention is not limited to the above-mentioned embodiments, and all technical solutions under the idea of the present invention belong to the protection category of the present invention. It should be pointed out that for those skilled in the art, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims (4)

1. accelerate Cascading Methods for HEVC coding rate, it is characterized in that described method comprises the steps:

Step 1, carries out precoding to incoming frame, obtains optimum macroblock partitions result;

First H.264 encoder is utilized to carry out to incoming frame the optimum macroblock partitions result that precoding obtains this frame, in precoding process, the framework of encoder adopts hybrid encoding frame H.264/AVC, input parameter is the reconstruction frames of present frame and HEVC, and the reconstruction frames of HEVC will by the reference frame as present frame;

Step 2, is mapped as HEVC set of patterns by optimum macroblock partitions H.264/AVC;

The optimum by the method mapped the result of the optimum macroblock partitions H.264/AVC obtained being mapped as HEVC may set of patterns; For the CU of different size, different mapped modes is taked to obtain the set of patterns of HEVC respectively respectively:

A.64 × 64 layers

For the CU of 64 × 64, only use the predictive mode of SKIP and 2N × 2N;

B.32x32 layer

For 32 × 32 layers, CU is divided into five kinds of different types according to the type of the MB of four in this CU, for the CU of every type, designs different mapping relations to ensure that the set mapping out meets most the feature of this CU;

C.16 × 16 and 8 × 8 layers

For the CU of 16 × 16 and 8 × 8, take different mapping relations respectively to obtain optimum PU set of patterns;

Step 3, use HEVC encoder to carry out the PU set of patterns of the different size CU of Ergodic Maps, utilance aberration optimizing obtains best CU and PU pattern.

2. a kind of Cascading Methods accelerating HEVC coding according to claim 1, is characterized in that in the B described in step 2: the pattern of five kinds of CU is as follows with the mapping of its correspondence:

A), when to comprise 3 or 4 patterns in current C U be the MB of Inter16 × 16, Inter2N × 2N pattern of current C U is only detected;

B), when the pattern of two MB in the left side is Inter16 × 16 in current C U, Inter2N × 2N and Inter2N × N pattern of current C U is only detected;

C), when the pattern of two MB in the right is Inter16 × 16 in current C U, Inter2N × 2N and InterN × 2N pattern of current C U is only detected;

When the pattern of four MB d) in current C U is Intra16 × 16, only detect Intra2N × 2N pattern of current C U;

E) for other situations, all patterns all will be detected to judge the optimal predictive mode of current C U.

3. a kind of Cascading Methods accelerating HEVC coding according to claim 1 and 2, is characterized in that in C described in step 2:

For the CU of 16 × 16 and 8 × 8, take different mapping relations to obtain optimal PU collection, concrete mapped mode is:

A) for 16 × 16 CU, SKIP mode map is H.264/AVC the SKIP/Merge pattern of HEVC; 16 × 16 mode map are 2N × 2N and SKIP/Merge pattern; H.264/AVC 16 × 8 and 8 × 16 respectively SKIP/Merge, 2N × 2N of corresponding HEVC and left and right/rectangle and the asymmetric partition modes up and down of correspondence; Will detect all patterns except Intra and N × N pattern for 8 × 8, HEVC, for the Intra pattern in H.264/AVC, HEVC will detect all patterns;

B) for 8 × 8 CU, SKIP mode map is H.264/AVC the SKIP/Merge pattern of HEVC; 16 × 16 mode map are 2N × 2N and SKIP/Merge pattern; H.264/AVC 16 × 8 and 8 × 16 respectively SKIP/Merge, 2N × 2N of corresponding HEVC and left and right/rectangle and the asymmetric partition modes up and down of correspondence; Will detect all patterns except Intra pattern for 8 × 8, HEVC, for the Intra pattern in H.264/AVC, HEVC will detect all patterns.

4. a kind of Cascading Methods accelerating HEVC coding according to any one of claim 1-3, when it is characterized in that described method is encoded for HEVC, effectively can utilize macroblock partitions result H.264/AVC, and macroblock partitions result is mapped as the possible set of patterns of the CU model selection that can be used for HEVC, reduce calculative CU and PU pattern, reach and accelerate HEVC coding object.