CN102665079A - Adaptive fast intra prediction mode decision for high efficiency video coding (HEVC) - Google Patents

Abstract

The invention discloses an adaptive intra prediction mode decision method for high efficiency video coding (HEVC). According to the method, optimal mode selection methods of rough mode decision (RMD), most probable mode (MPM) and rate distortion optimization (RDO) in the HEVC are used, a texture-based additional mode selection process is added after RMD, the quantity of candidate modes is reduced to 2-5 for prediction blocks with the size of 4*4 and 8*8, the quantity of the candidate modes is reduced to 2-5 for prediction blocks with the size of 16*16 and 32*32, so that the quantity of the candidate modes participating in the RDO mode selection is obviously reduced.

Description

Adaptive Fast Intra Prediction Mode Decision for HEVC

joint research

This application is jointly researched by North China University of Technology and Information Institute of Beijing Jiaotong University, and supported by the following funds: National Natural Science Foundation of China (No. Research Fund (No.2011JBM214).

technical field

The present invention relates to the field of image processing, and more specifically, to an adaptive fast intra-frame prediction mode decision method for intra-frame prediction mode selection in High Efficiency Video Coding (HEVC).

Background technique

In April 2010, the two major international video coding standard organizations VCEG and MPEG established the joint video compression group JCT-VC (Joint collaborative Team on Video Coding) to jointly develop the HEVC standard, also known as H.265. The main goal of the HEVC standard is to achieve a substantial improvement in coding efficiency compared with the previous generation standard H.264/AVC, especially for high-resolution video sequences. Its goal is to reduce the bit rate to 50% of the H.264 standard at the same video quality (PSNR).

At the current stage, HEVC still uses the hybrid coding framework adopted by H.264. Inter- and intra-frame predictive coding: De-correlation between temporal and spatial domains. Transform coding: Transform coding is performed on the residual to remove spatial correlation. Entropy coding: remove statistical redundancy. HEVC will focus on researching new coding tools or technologies within the framework of hybrid coding to improve video compression efficiency.

At present, many new coding features that have been proposed in the discussion organized by JCT-VC may be added to the HEVC standard. The specific documents of each discussion can be obtained from http://wftp3.itu.int .

The size of the encoding block will be up to 64x64, which is mainly for the application of high-definition video compression encoding. DCT transform coding will break through 8x8, and the maximum can reach 32x32. For intra prediction, the direction of prediction is more refined, and there are as many as 35 intra prediction modes, of which there are 17 prediction units with sizes of 4x4, 8x8, 16x16, 32x32 and 64x64 respectively. , 35, 35, 35 and 5 prediction modes are available, wherein, Fig. 2 shows the situation of 35 intra prediction modes, which will make the intra prediction more accurate and reduce redundancy more efficiently. Inter-frame prediction uses more tapped filters and 1/4 pixel precision during interpolation to improve the accuracy of inter-frame prediction. In terms of entropy coding, the more adaptable CABAC and the low-complexity LCEC are used.

In terms of intra-frame prediction, since up to 35 intra-frame prediction modes are used, when performing intra-frame coding, if you directly use rate-distortion optimization (Rate-Distortion optimization, RDO) for the 35 intra-frame prediction modes to judge If the optimal prediction mode is not used, the encoder will not be able to bear the computational load of RDO for all prediction directions. In order to reduce the computational load of the encoder when performing RDO operations, HM4.0 (HEVC test model 4.0) uses the Hadamard transform method to select several better candidate modes through a rough mode decision (Rough Mode Decision, RMD) . Then, RDO operation is performed on these several better candidate modes to obtain the best mode.

In the JCTVC-C207 proposal that JCT-VC has adopted, the minimum absolute sum of Hadamard Transformed coefficients of residual signal (HSAD) and the calculation process of the number of mode bits are added to the rough mode decision (RMD) algorithm. In the proposal JCTVC-D283, the joint direction intra prediction algorithm is simplified, and the most probable mode (MPM) algorithm is added to extract the highest probability mode (most probable mode, MPM) algorithm, that is, the MPM algorithm is applied after the RMD, so as to ensure that in the candidate mode set for RDO operation Always include the MPM. The proposal JCTVC-C218 simplifies the redundant process of dividing CUs into PUs. These simplified algorithms reduce the complexity of intra prediction, but these solutions are the most basic solutions without considering the nature of blocks.

This application mainly refers to the following technical documents, which are all JCT-VC and can be obtained directly from http://wftp3.itu.int :

[1] J.H.Min, "Unification of the directional intra prediction methods inTMuC," JCTVC-B100, July, 2010.

[2] K.McCann et.al., "Samsung's response to the call for proposals on video compression technology," JCTVC-A124, April, 2010.

[3] K.Ugur, K.R.Andersson and A.Fuldseth, "Description of video codingtechnology proposal by Tandberg, Nokia, Ericsson," JCTVC-A119, April, 2010.

[4] Y.Piao, J.H.Min, J.Chen, "Encoder improvement of unified intraprediction," JCTVC-C207, Oct., 2010.

[5] L.Zhao, et.al., "Further Encoder Improvement of intra mode decision," JCTVC-D283, Jan., 2011.

[6] J.Kim, Y.Jeon, B.Jeon, "Encoding complexity reduction for intraprediction by disabling NxN partition," JCTVC-C218, Oct., 2010.

[7]B.Bross, W.J.Han, J.R.Ohm, G.J.Sullivan, T.Wiegand, "WD4: Working Draft 4 of High-Efficiency Video Coding," JCTVC-F803, July, 2011.

[8] JCT-VC, HEVC reference software "HM-4.0" [online] http://hevc.hhi.fraunhofer.de/svn/svn_HEVCSoftware/tages

[9] Y.H.Tan, C.Yeo, H.L.Tan, Z.Li, "RQT depth selection," JCTVC-E104, Mar., 2011.

[10]K.Panusopone, K.Chono, Y.H.Tan, M.Zhou, "Evaluation of RQT inHM and related TU representation," JCTVC-E365, Mar., 2011.

[11] K.McCann, S.Sekiguci, B.Bross, W.J.Han, "HM4: HEVC TestModel 4 Encoder Description," JCTVC-F802, July, 2011.

[12]B.Bross, "Common test conditions and software reference configurations," JCTVC-F900, July, 2011.

但是候选模式的数量仍然有下降的空间。In terms of intra prediction, although several candidate mode selection algorithms have been proposed to reduce the number of modes for final RDO while satisfying BD-rate (BD-rate)

But the number of candidate patterns still has room to drop.

Contents of the invention

According to one aspect, there is provided an adaptive intra-frame (Intra) prediction mode decision method for high-efficiency video coding (HEVC), wherein the method is to perform a prediction mode for each prediction block (PU) in the current frame For decision-making, the method includes:

(1) Use Rough Mode Decision (RMD) to determine a candidate mode set, which contains N candidate modes with the minimum cost in the mode set, and the N candidate modes have been sorted in ascending order according to the cost sorting, where N=8 for prediction blocks of size 4*4 and 8*8, and N=3 for prediction blocks of size 16*16 and 32*32;

(2) When the prediction block size used is 4*4 or 8*8,

Judging whether the first two angle patterns in the set are adjacent: if they are adjacent, keep the first two angle patterns and the angle pattern closest to the first angle pattern among the other angle patterns in the set, if not , then only the first two angle modes are retained;

Judging whether there is a DC mode or a Planar mode in the first 4 modes in the set, and if so, retaining the DC mode and the Planar mode in the set; or

When the prediction block size used is 16*16 and 32*32,

Judging whether the first pattern in the set is a DC pattern or a Planar pattern, if so, retaining the DC pattern and the Planar pattern in the set, and directly proceeding to step (4);

(3) using the highest probability mode (MPM) algorithm, wherein the two highest probability modes are added to the set;

(4) Apply rate-distortion optimization (RDO) to each mode in the set to obtain the best intra prediction mode.

According to another aspect, there is provided a video encoder for High Efficiency Video Coding (HEVC), comprising: an intra prediction mode decider that determines an optimal intra prediction mode, and assigns the optimal intra prediction mode a mode input intra predictor; and an intra predictor that performs intra prediction based on the best intra prediction mode, wherein the intra predictor is for each prediction block (PU) in the current frame performing prediction mode decision-making, wherein the intra-frame prediction mode decider is configured to:

(1) Use Rough Mode Decision (RMD) to determine a candidate mode set, which contains N candidate modes with the minimum cost in the mode set, and the N candidate modes have been sorted in ascending order according to the cost sorting, where N=8 for prediction blocks of size 4*4 and 8*8, and N=3 for prediction blocks of size 16*16 and 32*32;

(2) When the prediction block size used is 4*4 or 8*8,

Judging whether the first two angle patterns in the set are adjacent: if they are adjacent, keep the first two angle patterns and the angle pattern closest to the first angle pattern among the other angle patterns in the set, if not , then only the first two angle modes are retained;

Judging whether there is a DC mode or a Planar mode in the first 4 modes in the set, and if so, retaining the DC mode and the Planar mode in the set; or

When the prediction block size used is 16*16 and 32*32,

Judging whether the first pattern in the set is a DC pattern or a Planar pattern, if so, retaining the DC pattern and the Planar pattern in the set, and directly proceeding to step (4);

(3) using the highest probability mode (MPM) algorithm, wherein the two highest probability modes are added to the set;

(4) Apply rate-distortion optimization (RDO) to each mode in the set to obtain the best intra prediction mode.

A device corresponding to the above method and a video encoder for performing the above method are also disclosed.

Description of drawings

Figure 1 shows an embodiment of an HEVC encoder block diagram.

Figure 2 shows up to 35 prediction modes used in HEVC.

Fig. 3 shows a flowchart of intra prediction mode selection according to an embodiment of the present invention.

Detailed ways

Various aspects are now described with reference to the figures. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of one or more aspects. It may be evident, however, that such aspects can be practiced without these specific details.

As used in this application, the terms "component", "module", "system" and the like are intended to refer to a computer-related entity such as, but not limited to, hardware, firmware, a combination of hardware and software, software, Or software in execution. For example, a component may be, but is not limited to being limited to, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. For example, both an application running on a computing device and the computing device can be components. One or more components can reside within a process and/or thread of execution and a component can be localized on one computer and/or distributed between two or more computers. In addition, these components can execute from various computer readable media having various data structures stored thereon. Components can communicate by means of local and/or remote processes, such as from signals having one or more data packets, for example, from interacting with another component in a local system, a distributed system, and/or with another component in a distributed system by means of a signal. Data of a component on a network such as the Internet that interacts with other systems by means of signals.

FIG. 1 shows a general block diagram of a video encoder implemented by High Efficiency Video Coding (HEVC). The encoder architecture of HEVC is roughly the same as the encoder architecture used by H.264, mainly for further research and improvement on the algorithms used in each module, especially for high-resolution video sequences. The goal of the improvement is Under the same video quality (PSNR), the bit rate is reduced to 50% of the H.264 standard.

Since the encoder architecture of HEVC is roughly the same as the encoder architecture used by H.264, the present invention will not be confused, and the overall architecture in Figure 1 will not be described in this application, but only focus on the intra prediction mode for HEVC selection and intra prediction.

Intra-frame prediction is a new intra-frame coding method first adopted in the H.264/AVC video coding standard, and its purpose is to remove spatial redundancy between the current block and adjacent blocks. In intra prediction, a predicted block P is formed from previously coded and reconstructed blocks in the current slice. The prediction block P is subtracted from the current block to generate a residual block. After block transformation and quantization, the residual block generates a quantized transformation coefficient X, which is then entropy coded and combined with other information required by the decoder side (for example, prediction mode quantization parameters, motion vectors, etc.) together form a compressed bit stream for transmission and decoding. In the H.264/AVC video coding standard, there are 9 prediction modes including direct current (DC) for 4*4 blocks, and there are 9 prediction modes including DC and planar for 8*8 and 16*16 blocks. ) including 4 modes. When performing intra prediction mode selection, the mode with the best match (for example, with the smallest rate-distortion (RD)) to the current block is selected from the supported prediction modes.

In HEVC, especially in HM4.0 (HEVC test model 4.0) (reference [11]), for intra prediction, the sizes are 4*4, 8*8, 16*16, 32*32 and 64 *64 prediction block (predication unit), and the number of intra prediction modes supported according to the size of the prediction block is shown in Table 1, and the possible prediction modes are shown in Figure 2. As shown in FIG. 2 , there are up to 35 prediction modes including 33 angle intra prediction mode directions, DC (direct current) mode and Planar (planar) mode.

predict block size Number of intra prediction modes 4 17 8 35 16 35 32 35 64 5

Table 1

In order to reduce the number of complex RD operations for intra-frame prediction, HEVC's intra-frame prediction mode applies Rough Mode Decision (Rough Mode Decision, RMD) and extracts MPM (most probable mode, MPM) method to obtain a streamlined best candidate mode set, Thereby reducing the number of RD operations. The purpose of RMD is to roughly guess the best intra prediction mode, and its selection process takes into account the HSAD operation value and the number of mode bits, but RMD does not explore the correlation between extracted modes. Therefore, analyzing the regularity between extracted modes can further reduce the complexity of joint intra prediction.

RMD algorithm

In order to reduce the computational load of the encoder when performing RDO operations, HM4.0 first uses the Hadamard transform method, and selects several better candidate modes through RMD. Then, RDO operation is performed on these several better candidate modes to obtain the best mode.

Specifically, in RMD, the following formulas are used to obtain 8 candidate modes for prediction blocks of 4*4 and 8*8 sizes, and 3 candidate modes for prediction blocks of 16*16 and 32*32 sizes:

J=HSAD(Res _mode )+λ·B _mode , (1)

Among them, J is the cost, _mode is all intra-frame prediction modes with up to 35 types specified by HEVC, HSAD (Res _mode ) represents the absolute sum of Hadamard transform coefficients of the residual signal, B _mode represents the consumption value of the number of mode bits, and λ represents Lagrang daily volume.

Therefore, the RMD algorithm is to use formula (1) to select 3 (for 16*16 and 32*32 sizes) or 8 (for 4* 4 and 8*8 sizes) candidate patterns.

More specifically, in the RMD algorithm, for each of all supported prediction modes, the associated cost J is calculated, and each prediction mode is sorted according to the descending order of the associated cost J, and the top 3 (for 16*16 and 32*32 sizes) or 8 (for 4*4 and 8*8 sizes) prediction modes are added to the candidate mode set as candidate modes.

Since the complexity of formula (1) is much smaller than that of RDO in terms of implementation, before using the RDO algorithm to determine the best prediction mode, using the RMD algorithm to perform a pre-selection among all supported prediction modes can significantly improve Reduced intra prediction time.

RDO Algorithm

RDO is an algorithm commonly used in intra-frame prediction and inter-frame prediction in video coding, which is usually based on the minimum mean square error (SSE: Sum of Square Error) to determine the best matching block.

In order not to confuse the key concepts in this article, it will not be described in detail.

MPM algorithm

The MPM algorithm is an algorithm that has been proposed in HM3.0, and has been further improved in JCTVC-D283.

There are various ways of determining the MPM in various proposals related to the MPM algorithm. This article uses the MPM algorithm in JCTVC-D283. Specifically, after using the RMD algorithm to obtain the condensed best mode candidate set, the best mode of the left prediction block of the current prediction block and the best mode of the upper prediction block are directly used as the two highest probability modes (MPM ), and judge whether the MPM is included in the candidate set, if not included in the set, then join the two MPMs. In other words, regardless of the mode selection algorithm before RDO, the best mode of the left prediction block and the best mode of the upper prediction block of the current prediction block are added.

In HM4.0, there are two steps in joint intra prediction to extract the intra prediction mode. In the first step, all the patterns are selected through HSAD operation to select the best pattern candidate set (RMD). In the best mode candidate set, the number of modes varies with the size of the PU. The 4*4 and 8*8 size block sets contain 8 candidate modes, while the 16*16 and 32*32 size block sets contain 3. Then, MPMs extracted from neighboring blocks are also added to the set. In the second step, each mode in the best mode set is subjected to RD optimization calculation (RDO), and the mode with the smallest RD consumption value is regarded as the best mode for intra-frame prediction.

After our observation and analysis, the smallest HSAD pattern and the second smallest HSAD pattern selected in the RMD process characterize the texture characteristics of the block: the adjacent smallest HSAD pattern and the second smallest HSAD pattern represent the block is a regular texture area; non-adjacent The smallest HSAD mode and the second smallest HSAD mode, DC mode or Planar mode represent blocks that are irregularly textured regions. In our proposed algorithm, we use this feature to adaptively reduce the number of RD operations in the best mode set to reduce the complexity of HEVC intra prediction.

Adaptive Candidate Pattern Selection Algorithm

Referring to FIG. 3 , a flow chart of intra prediction mode selection according to an embodiment of the present invention is shown.

In step 201, for example, the RMD algorithm performs mode selection among all supported prediction modes. Specifically, use formula (1) to obtain 8 candidate modes for prediction blocks of 4*4 and 8*8 sizes (step 205), and obtain 3 candidate modes for prediction blocks of 16*16 and 32*32 sizes (step 207).

Specifically, for each of all prediction modes supported by the corresponding prediction block size, calculate the associated cost J, sort the prediction modes in descending order of the associated cost J, and select the first three (for 16*16 and 32*32 sizes) or 8 (for 4*4 and 8*8 sizes) prediction modes are added to the candidate mode set as candidate modes. Moreover, in the candidate mode set, each candidate prediction mode is also sorted in descending order of the cost J.

For a prediction block with a scale of 64*64, the present invention does not select an RMD mode for it.

For prediction blocks with sizes of 4*4 and 8*8, in step 209 , it is judged whether the first two angle modes in the candidate mode set are adjacent: if they are adjacent, proceed to step 211 . In step 211, the first two angle patterns and the nearest neighbor angle patterns to the first angle pattern among other angle patterns in the set are kept in the candidate pattern set, that is, only three angle patterns are kept. If it is judged in step 209 that they are not adjacent, only the first two angle patterns are reserved in the candidate pattern set (step 213), that is, only two angle patterns are reserved.

The process continues to step 215 . In step 215, it is judged whether there is a DC (direct current) mode or a Planar (plane) mode in the first 4 candidate modes in the candidate mode set, and if it exists, the DC mode and the Planar mode are reserved in the candidate mode set, and the The process proceeds to step 223 . If it is determined in step 215 that there is no DC mode or Planar mode in the first 4 candidate modes, proceed directly to step 223 .

For prediction blocks of size 16*16 and 32*32, in step 209, it is judged whether the first mode in the candidate mode set is DC mode or Planar mode, and if so, the DC mode and all Planar mode described above, and proceed to step 225 directly. If it is determined in step 209 that the first mode in the candidate mode set is not a DC mode or a Planar mode, proceed to step 223 .

In step 223, using the MPM algorithm, the best mode of the left prediction block of the current prediction block and the best mode of the upper prediction block are directly added to the candidate mode set as two highest probability modes (MPMs). Of course, if the two MPMs are already included in the candidate mode set, the adding operation can be skipped.

In step 225, rate-distortion optimization (RDO) is applied to each mode in the set of candidate modes to obtain the best intra prediction mode. It is known to those skilled in the art that the best intra prediction mode is the prediction mode that obtains the smallest rate-distortion (RD).

After determining the best intra-frame prediction mode of the current block, the encoder can perform intra-frame prediction on the current frame according to the best intra-frame prediction mode, as described above, and finally obtain an encoded bit stream. The specific process of performing intra prediction does not belong to the focus of the present invention, so it will not be repeated here.

In the method shown in Figure 3, for prediction blocks of size 4*4 and 8*8, the number of candidate modes is reduced to 2-5, while for prediction blocks of size 16*16 and 32*32, the number of candidate modes is reduced to The number is reduced to 2-5, thereby significantly reducing the number of candidate modes participating in RDO mode selection.

The comparison of the experimental results of the present invention is based on the HM4.0 code platform, and the test environment is set according to the requirements of the proposal JCTVC-F900. Since our experiments focus on intra-frame prediction, we employ full I-frame encoding for both HE and LC test conditions. Under the HE and LC test environments, Table 2 and Table 3 respectively represent the BD-rate and encoding time of the three components of Y, U, and V compared with the default HM4.0. According to the experimental results, the scheme in this paper reduces the encoding time by 15% and 20% when the BD-rate only increases by 0.64% and 1.05%.

Table 2. BD-rates (%) and running time under high-efficiency full I-frame test conditions

Table 3. BD-rates (%) and running time under low-complexity full I-frame test conditions

The adaptive fast intra-frame prediction mode decision-making method disclosed in the present invention can be realized by software, hardware, firmware and the like.

When implemented in hardware, video encoders can be implemented using general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gates, or transistor logic devices, discrete hardware components, or any combination thereof designed to perform the functions described herein, may be implemented or performed. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, eg, a DSP and a microprocessor, multiple microprocessors, one or more microprocessors with a DSP core, or any other such architecture. Additionally, at least one processor may comprise one or more modules operable to perform one or more of the steps and/or operations described above.

When a video encoder is implemented with a hardware circuit such as an ASIC, FPGA, it may include various circuit blocks configured to perform various functions. Those skilled in the art can design and implement these circuits in various ways according to various constraints imposed on the entire system, so as to realize various functions disclosed in the present invention.

Although the foregoing disclosures discuss exemplary aspects and/or embodiments, it should be noted that many changes and/or changes may be made therein without departing from the scope of the described aspects and/or embodiments as defined by the claims. Revise. Furthermore, although elements of the described aspects and/or embodiments are described or claimed in the singular, the plural is contemplated unless limitation to the singular is explicitly stated. In addition, all or part of any aspect and/or embodiment can be used in combination with all or part of any other aspect and/or embodiment, unless a difference is indicated.

Claims (10)

1. An adaptive Intra (Intra) prediction mode decision method for High Efficiency Video Coding (HEVC), wherein the method performs a prediction mode decision for each prediction block (PU) in a current frame, the method comprising:

(1) determining a set of candidate patterns using a coarse pattern decision (RMD), in which the N candidate patterns of the set of patterns having the smallest cost are included and have been sorted in ascending order by cost, wherein N is 8 for prediction blocks of sizes 4x4 and 8x8 and N is 3 for prediction blocks of sizes 16x16 and 32x 32;

(2) when the prediction block size used is 4x4 or 8x8,

judging whether the first 2 angle modes in the set are adjacent: if the angle modes are adjacent, the angle mode which is the nearest neighbor of the first 2 angle modes and other angle modes in the set and is the nearest neighbor of the first 1 angle mode is reserved, and if the angle modes are not adjacent, only the first two angle modes are reserved;

judging whether a DC mode or a Planar mode exists in the first 4 modes in the set, if so, reserving the DC mode and the Planar mode in the set; or

When the prediction block sizes used are 16x16 and 32x32,

judging whether the 1 st mode in the set is a DC mode or a Planar mode, if so, reserving the DC mode and the Planar mode in the set, and directly proceeding to the step (4);

(3) using a highest probability mode (MPM) algorithm, wherein two highest probability modes are added to the set;

(4) rate Distortion Optimization (RDO) is applied to each mode in the set to obtain the best intra prediction mode.

2. The method of claim 1, wherein if the size of the prediction block is 64x64, the step (4) is directly performed on the 3 prediction modes associated with the prediction block of 64x64 size without performing the steps (1) - (3).

3. The method of claim 1, wherein the cost is based on:

J＝HSAD(Res_mode)+λ·B_mode，

wherein,_modeall intra prediction modes, HSAD (Res) with a maximum of 35, specified for HEVC_mode) Representing the absolute sum of the Hadamard transform coefficients of the residual signal, B_modeRepresents the consumption value of the number of mode bits and λ represents the lagrangian constant.

4. A video encoder for High Efficiency Video Coding (HEVC), comprising:

an intra prediction mode decider which determines an optimal intra prediction mode and inputs the optimal intra prediction mode to an intra predictor; and

an intra predictor that performs intra prediction based on the optimal intra prediction mode, wherein the intra predictor performs a prediction mode decision for each prediction block (PU) in a current frame,

wherein the intra prediction mode decider is configured to:

(1) determining a set of candidate patterns using a coarse pattern decision (RMD), in which the N candidate patterns of the set of patterns having the smallest cost are included and have been sorted in ascending order by cost, wherein N is 8 for prediction blocks of sizes 4x4 and 8x8 and N is 3 for prediction blocks of sizes 16x16 and 32x 32;

(2) when the prediction block size used is 4x4 or 8x8,

judging whether the first 2 angle modes in the set are adjacent: if the angle modes are adjacent, the angle mode which is the nearest neighbor of the first 2 angle modes and other angle modes in the set and is the nearest neighbor of the first 1 angle mode is reserved, and if the angle modes are not adjacent, only the first two angle modes are reserved;

judging whether a DC mode or a Planar mode exists in the first 4 modes in the set, if so, reserving the DC mode and the Planar mode in the set; or

When the prediction block sizes used are 16x16 and 32x32,

judging whether the 1 st mode in the set is a DC mode or a Planar mode, if so, reserving the DC mode and the Planar mode in the set, and directly proceeding to the step (4);

(3) using a highest probability mode (MPM) algorithm, wherein two highest probability modes are added to the set;

(4) applying Rate Distortion Optimization (RDO) to each mode in the set to obtain the best intra prediction mode.

5. The intra prediction mode decider as claimed in claim 4, wherein if the size of the prediction block is 64x64, the step (4) is directly performed on the 3 prediction modes associated with the prediction block of 64x64 size without performing the steps (1) - (3).

6. The intra prediction mode decision maker as claimed in claim 4, wherein the cost is based on:

J＝HSAD(Res_mode)+λ·B_mode，

wherein,_modeall intra prediction modes, HSAD (Res) with a maximum of 35, specified for HEVC_mode) Representing the absolute sum of the Hadamard transform coefficients of the residual signal, B_modeRepresents the consumption value of the number of mode bits and λ represents the lagrangian constant.

7. An adaptive Intra (Intra) prediction mode decision device for High Efficiency Video Coding (HEVC), the device performing a prediction mode decision for each prediction block (PU) in a current frame, the device comprising:

-a module (1) for determining a set of candidate patterns using a coarse pattern decision (RMD), in which the N candidate patterns with the smallest cost of the set of patterns are included and which have been sorted in ascending order according to cost, wherein N is 8 for prediction blocks of size 4x4 and 8x8 and N is 3 for prediction blocks of size 16x16 and 32x 32;

a module (2) for determining, when the prediction block size used is 4x4 or 8x8,

judging whether the first 2 angle modes in the set are adjacent: if the angle modes are adjacent, the angle mode which is the nearest neighbor of the first 2 angle modes and other angle modes in the set and is the nearest neighbor of the first 1 angle mode is reserved, and if the angle modes are not adjacent, only the first two angle modes are reserved;

judging whether a DC mode or a Planar mode exists in the first 4 modes in the set, if so, reserving the DC mode and the Planar mode in the set; or

When the prediction block sizes used are 16x16 and 32x32,

judging whether the 1 st mode in the set is a DC mode or a Planar mode, if so, reserving the DC mode and the Planar mode in the set, and directly proceeding to a module (4);

the module (3) uses a highest probability mode (MPM) algorithm, wherein the two highest probability modes are added to the set;

a module (4) applies Rate Distortion Optimization (RDO) to each mode in the set to obtain the best intra prediction mode.

8. The apparatus of claim 7, wherein if the size of the prediction block is 64x64, the step (4) is directly performed on the 3 prediction modes associated with the prediction block of 64x64 size without performing the steps (1) - (3).

9. The intra prediction mode decision maker as claimed in claim 7, wherein the cost is based on:

J＝HSAD(Res_mode)+λ·B_mode，

wherein,_modeall intra prediction modes, HSAD (Res) with a maximum of 35, specified for HEVC_mode) Representing the absolute sum of the Hadamard transform coefficients of the residual signal, B_modeRepresents the consumption value of the number of mode bits and λ represents the lagrangian constant.

10. A video encoder for performing the method of claims 1-3.

Images