CN100348051C - An enhanced in-frame predictive mode coding method - Google Patents

An enhanced in-frame predictive mode coding method Download PDF

Info

Publication number
CN100348051C
CN100348051C CN 200510018489 CN200510018489A CN100348051C CN 100348051 C CN100348051 C CN 100348051C CN 200510018489 CN200510018489 CN 200510018489 CN 200510018489 A CN200510018489 A CN 200510018489A CN 100348051 C CN100348051 C CN 100348051C
Authority
CN
China
Prior art keywords
prediction mode
block
prediction
current
mode
Prior art date
Application number
CN 200510018489
Other languages
Chinese (zh)
Other versions
CN1674680A (en
Inventor
朱光喜
戴声奎
刘文予
喻莉
Original Assignee
华中科技大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华中科技大学 filed Critical 华中科技大学
Priority to CN 200510018489 priority Critical patent/CN100348051C/en
Publication of CN1674680A publication Critical patent/CN1674680A/en
Application granted granted Critical
Publication of CN100348051C publication Critical patent/CN100348051C/en

Links

Abstract

本发明属于信号处理中的视频编解码领域,涉及一种增强型帧内预测模式编码方法。 The present invention belongs to the field of video signal encoding and decoding processing, directed to a method of intra prediction encoding modes enhanced. 其特点在于:利用相邻图像块的纹理的空间相关性,根据相邻块之间最佳预测模式的关系,统计确定视频的最可能预测模式表,从而提高帧内编码时预测精度。 Characterized in that: using a texture image blocks adjacent to the space correlation, the relationship between the optimum prediction mode of adjacent blocks, statistically determined most probable prediction mode table in the video, thereby increasing the intra prediction encoding precision. 本发明方法包括:统计相邻块之间最佳预测模式的关系,建立最可能预测模式查找表;在编码端,根据左邻块和上邻块的预测模式查表得出当前块的最可能预测模式,然后进行相应的帧内预测编码操作;在解码端,根据左邻块和上邻块的预测模式查表得出当前块的最可能预测模式,然后进行相应的帧内预测解码操作。 The method of the present invention comprises: the relationship between the optimum prediction mode of adjacent blocks statistical establish the most likely prediction mode table lookup; at the encoder, according to the left adjacent block and the prediction mode of block look-up table on the neighbor of a current block obtained most likely prediction mode and the corresponding prediction coding operation; at the decoding end, according to the left adjacent block and the prediction mode of block look-up table on the ortho derived most likely prediction mode of the current block, then the appropriate intra prediction decoding operation. 本发明所提出的方法简单易行,能有效增加帧内编码时预测精度,从而提高视频的编码效率。 The proposed method of the present invention is simple, can effectively increase the precision of intra prediction encoding, thereby improving the coding efficiency of the video.

Description

一种增强型帧内预测模式编码方法 Of an intra prediction mode encoding method Enhanced

技术领域 FIELD

本发明属于信号处理中的视频编解码领域,涉及一种增强型帧内预测模式编码方法。 The present invention belongs to the field of video signal encoding and decoding processing, directed to a method of intra prediction encoding modes enhanced.

背景技术 Background technique

随着通信技术与计算机技术的飞速发展,数字音视频信息的需求量与日俱增。 With the rapid development of communication technology and computer technology, the demand for digital audio and video information is growing. 由于原始视频数据量极大,为了节省信道传输带宽和存储空间,需要对其进行压缩编码。 Since the original amount of video data greatly, in order to save channel transmission bandwidth and storage space, it needs to be compression-encoded. 在视频编码技术的发展过程中,产生了一系列国际标准,如国际电信联盟(ITU-T)制定的H.261、H.263系列,国际标准化组织(ISO)制定的MPEG-1、MPEG-2、MPEG-4系列。 In the development process of video coding technology, it produces a series of international standards such as the International Telecommunication Union (ITU-T) developed H.261, H.263 series, the International Organization for Standardization (ISO) developed by MPEG-1, MPEG- 2, MPEG-4 series. 目前,ISO和ITU建立的联合视频工作组(JVT)共同制定了最新的视频编码标准H.264/MPEG-4 Part 10。 Currently, the Joint Video Team (JVT) ISO and ITU established jointly developed the latest video coding standard H.264 / MPEG-4 Part 10. 在国内,中国数字音视频编解码技术标准(AVS)工作组正在制定面向存储应用的AVS标准和面向移动通信的中国音视频编解码标准AVS-M。 In China, the China Audio Video Coding Standard (AVS) AVS working group is developing standards for storage applications and Chinese audio and video codec standard AVS-M for mobile communications.

由于视频数据在空间域上有较强的相关性,待处理象素块的值可由其邻近象素的值来预测。 Since the video data has a strong correlation in the spatial domain, the value of pixel block to be processed may be predicted by the value of adjacent pixels. 将实际值与预测值的残差进行编码可获得较大的压缩率。 The actual and predicted values ​​of the residual encoding can be obtained a larger compression ratio. H.264中采用基于块的空间预测方法对帧内图像编码。 The method using spatial prediction in H.264 intra blocks based image coding. 对于亮度分量,可采用4×4与16×16两种分块方式。 For the luminance component, can be 4 × 4 and 16 × 16 block two kinds of ways. 在帧内预测的4×4模式下,每个4×4块中的象素通过邻块中已经解码的17个象素进行预测。 In the intra 4 × 4 prediction mode, each pixel in the 4 × 4 block is predicted by a block already decoded o 17 pixels. 预测模式分为9种,在编码时应该为每个4×4块选择一种残差最小的最佳预测模式。 It is divided into nine kinds of prediction modes, when coding each be 4 × 4 block to select a minimum residual optimum prediction mode. 在编码端需要把最终采用的帧内预测模式的信息写进码流中,以便正确解码。 Encoding side information needed to use the intra prediction mode is written into the final code stream to be decoded correctly. 为了最大程度地压缩码字,H.264在编码端采用一个简单有效的策略:首先根据当前块的左邻块和上邻块的状态确定当前块最可能预测模式(mostProbableMode),如果上邻块A和左邻块B都在当前帧内,则最可能预测模式为A和B中较小的模式,否则最可能预测模式为均值(DC)预测模式;然后对每一种有效的模式计算代价,用代价最小的模式作为当前块的预测模式;最后把当前块的预测模式信息写进码流,为了节省码字,如果最终预测模式等于最可能预测模式,则简单地设置信息标志为1,否则把当前模式信息完整地写入码流中。 In order to maximize the compression codewords, H.264 uses a simple and effective strategy in the encoder: first determine the most probable prediction mode of a current block (MostProbableMode) The left adjacent block and the neighbor blocks in the current state of the block, the block if o a and B are left adjacent block in the current frame, the most likely prediction mode a and mode B, the smaller, or most likely prediction mode as mean (DC) mode prediction; then calculates a cost effective model for each , with minimal expense, as the mode of the current prediction mode of a block; the final prediction mode of the current block is written into the information stream, in order to save the code word, if the most probable final prediction mode is equal to prediction mode information flag is simply set to 1, otherwise, the current mode information is completely written in the code stream. 由此可见,最可能预测模式的预测越准确,码流的节省就越大。 Thus, most likely the more accurate prediction model, the greater the savings stream.

但实验表明,在很多情况下预测模式的取小原则与实际存在一定的误差,由于最可能预测模式值在最后模式比较时占有优势,使得最后的模式并非最佳模式,这样就不能最大程度地节省码流和提高预测精度。 But experiments show that, in many cases, take a small prediction mode and the actual principle there are some errors, most likely due to the prediction mode value advantage in the final comparison mode, so that the final model is not the best mode, so you can not maximize the save stream and improve prediction accuracy.

发明内容 SUMMARY

本发明的目的在于克服上述帧内预测模式编码方法的不足之处,提供一种增强型帧内预测模式编码方法。 Object of the present invention is to overcome the shortcomings of the above-described intra-prediction mode encoding method, providing an enhanced intra prediction mode encoding method. 该方法简单易行,能更好地提高压缩率和增加预测精度。 The method is simple, can better improve the compression ratio and the increase in prediction accuracy.

本发明提供的一种增强型帧内预测模式编码方法,首先建立最可能预测模式查找表;在编码端根据该查找表对帧内预测的每个块进行编码;在解码端根据该查找表对帧内预测的每个块进行解码;建立最可能预测模式查找表的步骤为:(A1)将测试序列对应的预测模式统计表stat[9][9][9]中设置初始值为0;(A2)对每个图像块进行帧内预测编码处理:首先在中国数字音视频编解码技术标准AVS或国际标准组织ISO和国际电信联盟ITU联合视频工作组制定的视频编码标准H.264规定的每一种预测模式下,设置率失真代价值变量的初始值为零,如果当前块是图像的上边缘块或左边缘块,选择均值DC预测模式作为其最可能预测模式,写入最可能预测模式查找表;否则进行下述处理:(A3)在AVS或H.264标准规定的每一种预测模式下计算其率失真代价值,选定率失真代价值最小的模式C An enhanced intra prediction mode encoding process of the present invention provides, firstly establish the most likely prediction mode table lookup; coding each block in the intra prediction encoding based on the lookup table end; decoding end based on the lookup table each block of the intra prediction decoding; look-up table step establishing the most likely prediction mode of: (A1) a test sequence corresponding to the prediction mode statistics stat [9] [9] [9] in the initial value is set to 0; (A2) intra prediction encoding process for each image block: firstly developed in China digital audio and video coding standard AVS ISO or international standards Organization and the joint video Team international Telecommunication Union ITU H.264 video coding standard defined in each prediction mode, the rate-distortion cost value set variable is initially zero, if the current block is a block edge or the left edge of the image block, selecting a DC prediction mode as the mean of the most likely prediction mode, writing the most probable prediction the rate-distortion cost values ​​calculated in each prediction mode (A3) or specified in the H.264 standard AVS, substituting the minimum value of the selected rate distortion model C: otherwise, the following processing; mode lookup table 当前块的预测模式;在当前块的左邻块的预测模式为L,其上邻块的预测模式为U时,将stat[L][U][C]中的数加1; Current prediction mode of a block; in a left neighbor block of a prediction mode of the current block is L, the ortho-mode prediction block is U, the stat [L] [U] Number of [C] is incremented by 1;

对所有帧内预测的块重复步骤(A2)至(A3)操作,即得到该序列的预测模式统计表stat[9][9][9];(A4)重复上述步骤(A1)至(A3)对典型测试序列进行统计,每个测试序列得到一张预测模式统计表,将各统计表的对应项相加得预测模式统计总表stat_all[9][9][9];(A5)分别令左邻块预测模式L和上邻块预测模式U为0至8的所有整数,在每种L和U的组合下,按以下方法确定当前块最可能预测模式:对每个给定的L和U,找出stat_all[L][U][i]中的最大值stat_all[L][U][C],其中C就是该组合下当前块的最可能预测模式,,I为1到8的任意整数,将C写入最可能预测模式查找表的mpm[L][U];对帧内预测的每个块进行编码的步骤为:(B1)根据上邻块和左邻块的预测模式,查上述最可能预测模式查找表得出当前块的最可能预测模式:设当前块的左邻块的预测模式为L、上邻块的预测模式U,如果当前图像块 Repeat steps for all of the intra prediction block of (A2) to (A3) operation, i.e., to obtain the prediction mode of the sequence tables stat [9] [9] [9]; (A4) repeating the above steps (A1) to (A3 ) typical test sequence statistics, each test sequence to obtain a prediction mode tables, each corresponding term statistics obtained by adding the total statistical prediction mode table stat_all [9] [9] [9]; (A5), respectively so that the left adjacent block prediction mode L and the prediction mode of block U o for all integers from 0 to 8, in each combination of L and U, determining the most probable prediction mode of a current block as follows: for each given L and U, find stat_all [L] [U] [i] is the maximum stat_all [L] [U] [C], where C is the most likely prediction mode of the current block is a combination of 1-8 ,, I is any integer, the most likely prediction mode C is written in the lookup table mpm [L] [U]; step for each block of intra prediction encoding of: (B1) the upper adjacent block and the left block prediction o mode, the above-described most likely prediction mode search lookup tables stars most probable prediction mode of the current block: set the prediction mode of the current block is a left neighbor block is L, the ortho-mode prediction block U, if the current block 上边缘块或左边缘块,则Cp=DC;否则当前块的最可能预测模式Cp=mpm[L][U];(B2)计算当前块在每一种预测模式下的代价值,选择最佳的预测模式:分别在AVS标准或H.264标准规定的9种可能的预测模式下,设置率失真代价值的初始值,然后计算当前块的率失真代价值;选定率失真代价值最小的模式为当前块的预测模式C;(B3)以选定的预测模式C对当前块编码:首先计算当前块的原始象素值和该预测模式C下的预测值之间的差值,得到当前块的残差图像;再对残差图像块进行整数变换和量化;对计算量化后的变换系数进行熵编码;(B4)将预测模式信息写入码流:为每个4×4块设定标志位f,如果Cp=C,则令f=1,否则f=0,将标志位f用1比特写入码流;用变量R表示当前要写入的模式信息:如果当前块的预测模式C小于最可能预测模式Cp,则R=C,否则R=C-1,将R对应 Block left edge or the upper edge of the block, then the Cp = DC; otherwise, most likely prediction mode of a current block Cp = mpm [L] [U]; (B2) to blocks of the current value is calculated in each prediction mode, to choose the most good prediction mode: in each of the 9 possible prediction modes specified in the AVS standard or H.264 standard, the rate-distortion cost value set initial value, and then calculates rate-distortion cost value of a current block; selected minimum rate-distortion cost value model for the prediction mode of the current block C; (B3) in the selected prediction mode for encoding the current block C: first, calculate the difference between the original pixel values ​​of the current block and the prediction mode value C, giving a residual image of the current block; and then the residual image block integer transform and quantization; calculated on the quantized transform coefficient entropy encoding; (B4) prediction mode information of the code stream is written: for each 4 × 4 block is provided set flag f, if Cp = C, then let f = 1, otherwise f = 0, the flag f is written with a bit stream; R represents a variable by the current mode information to be written: If a prediction of the current block mode C is less than the most likely prediction mode Cp, then R = C, or R = C-1, corresponding to the R 三位二进制数写入码流;(B5)将熵编码后的信息写入码流;对帧内预测的每个4×4块进行解码的步骤为:(C1)从码流中读取图像块的帧内编码信息,包括标志位f和模式信息R;(C2)确定当前块的最可能预测模式Cp:如果当前块为当前帧图像的左边缘块或上边缘块,则Cp为DC模式;否则设当前块的左邻块的预测模式为L、上邻块的预测模式U,则Cp为预测模式查找表中表项mpm[L][U]的值;(C3)确定当前图像块的帧内预测模式C:如果标志位f=1,则C=Cp;否则进行如下处理:如果模式信息R小于Cp,则C=R,否则C=R+1;(C4)重建当前图像块:根据预测模式C进行图像块象素的预测;对图像块数据进行熵解码得到变换系数,然后对变换系数进行反量化、反变换;最后得到重建的图像块。 Three write binary stream; (B5) to write information entropy coding code stream; each 4 × 4 block decoding step of intra prediction is: (C1) an image read from the code stream intra-coded block of information comprises mode information and flag f R; (C2) most likely determined that the current block is prediction mode Cp: if the current block to the left edge of the block of the current frame image or the upper edge of the block, the DC mode Cp ; otherwise, set the current block left adjacent block prediction mode is L, the upper adjacent block prediction modes U, the lookup value Cp mpm [L] [U] is a table of the prediction mode table entry; (a C3) determining a current image block intra prediction mode C: If the flag bit f = 1, then C = Cp; otherwise processed as follows: If the mode information R is less than Cp, the C = R, or C = R + 1; (C4) reconstructing the current image block : C according to a prediction mode a prediction image pixel block; data of the image block to obtain transform coefficients entropy decoding, the transform coefficients and inverse quantization, inverse transform; to give the final reconstructed image block.

本发明根据相邻块之间最佳预测模式的统计关系来估计最可能预测模式。 The present invention is to estimate the most likely prediction mode according to a statistical relationship between the optimum prediction mode of adjacent blocks. 其主要理论基础是相邻图像块的纹理具有空间相关性。 The main theory is based on the texture of adjacent image blocks having spatial correlation. 帧内预测所确定的最佳预测模式,其预测方向就是纹理信息最强的方向。 Optimum prediction mode determined by the intra prediction, the prediction direction is the direction of the strongest texture information. 所以,可以推论在H.264或AVS等编码系统中,相邻图像块的最佳预测模式也是相关的。 Therefore, H.264 or the like can be inferred in the AVS system, the optimum prediction mode of the adjacent image blocks are also relevant. 具体而言,本发明的主要优点是:其一、所提出的方法简单易行,不增加额外的计算复杂度。 Specifically, the main advantages of the present invention are: First method, proposed is simple, no additional computational complexity.

其二、所提出的方法能很好地增加预测精度、提高压缩率。 Second, the proposed method can well increase the prediction accuracy, increasing the compression ratio.

附图说明 BRIEF DESCRIPTION

图1为H.264和AVS中定义的一种8个方向预测模式及其标号示意图。 Figure 1 is an 8-directional prediction mode and reference numeral schematic H.264 and AVS defined.

具体实施方式 Detailed ways

本发明通过对大量测试序列的预测模式进行统计,得出在左邻块预测模式(leftmode)和上邻块预测模式(upmode)的各种组合下,当前块最佳预测模式的取值规律;将各种左邻块预测模式和上邻块预测模式组合下最可能的最佳模式作为最可能预测模式(mostProbableMode)写入一张固定的查找表中。 The present invention is by statistical prediction mode of a large number of test sequences, results in the left adjacent block prediction mode (leftmode), and various combinations of the prediction mode of block o (upmode), the current value of the optimum prediction mode block law; the various blocks left o o prediction mode and the prediction mode combination most likely block the best mode most likely prediction mode (MostProbableMode) writing a look-up table as fixed.

下面以AVS中4×4模式亮度分量的帧内预测为例,对本发明作进一步详细的说明。 AVS following luminance component to the intra 4 × 4 prediction mode as an example, the present invention will be further described in detail. AVS中定义的预测模式方向如图1所示。 AVS prediction mode defined in the direction as shown in FIG.

具体的编码过程如下:首先定义如下变量以便于算法描述:左邻块预测模式L(leftmode):记录当前块的左邻块选用的预测模式,其值为0到8的整数。 Specific encoding process as follows: First define the following variables for the algorithm described in: left adjacent block prediction mode L (leftmode): a recording mode of the left adjacent block of the current prediction block is selected, a value of an integer from 0-8.

上邻块预测模式U(upmode):记录当前块的上邻块选用的预测模式,其值为0到8的整数。 O the prediction mode of block U (upmode): o the recording mode prediction block of the current block is selected, a value of an integer from 0-8.

当前块预测模式C(currentmode):记录当前块选用的预测模式,其值为0到8的整数。 Prediction mode of the current block C (currentmode): a recording mode of the current prediction block is selected, a value of an integer from 0-8.

预测模式统计表stat[9][9][9]:每个测试序列对应一张预测模式统计表,用来记录该序列帧内预测模式的统计信息。 Prediction mode statistics stat [9] [9] [9]: Each test sequence corresponding to a prediction mode tables, records information of the sequence to an intra prediction mode. stat[L][U][C]表示,在该序列采用帧内预测的4×4图像块中,左邻块预测模式为L、上邻块预测模式为U、且当前预测模式为C的4×4块的个数。 stat [L] [U] [C] represented using the sequence intra 4 × 4 prediction block in the image, the left adjacent block prediction mode is L, the upper adjacent block prediction mode is U, and the current prediction mode is C the number of 4 × 4 block.

预测模式统计总表stat_all[9][9][9]:由所有测试序列的预测模式统计表对应项相加得到。 General statistical prediction mode table stat_all [9] [9] [9]: of all the prediction modes of the test sequence statistics obtained by adding the corresponding item.

最可能预测模式查找表mpm[9][9]:记录最可能的预测模式,其值为0到8的整数;当左邻块预测模式为L、上邻块预测模式为U时,当前块最可能的预测模式为mpm[L][U]。 The most likely prediction mode table lookup mpm [9] [9]: recording the most likely prediction mode, a value of an integer from 0 to 8; when the left adjacent block prediction mode is L, the upper adjacent block prediction mode is U, the current block the most likely prediction mode is mpm [L] [U].

标志位f:标志图像块最可能预测模式的选择是否准确,其值为二进制数0或1,在码流中占用1bit。 Flag f: symbol image block most likely to select the accuracy of the prediction mode, a value of binary 1 or 0, the code stream occupies in 1bit.

预测模式选择变量R(remaining_mode_sekector):当最可能预测模式的选择不准确时,用此变量表示当前块的预测模式信息,其值为0到7的整数,在码流中写入该变量的二进制数(对应000到111),占用3bit。 A prediction mode selection variable R (remaining_mode_sekector): When the most probable inaccurate prediction mode selected, the prediction mode represented by this variable current block information, which is an integer from 0 to 7, the variable is written in a binary code stream number (corresponding to 000 to 111), occupies 3bit.

一.编码前的准备工作:对大量已编码的典型序列预测模式的取值规律作统计实验,建立一般情况下的最可能预测模式查找表:步骤(A):统计每个测试序列的帧内预测模式取值规律:先将该序列对应的预测模式统计表stat[9][9][9]中所有项置初始值0,再逐个扫描采用帧内预测编码的4×4块(不包括图像帧的上边缘块和左边缘块),首先在AVS或H.264标准规定的每一种预测模式下,设置率失真代价值变量的初始值为零,在当前块的预测模式为C,其左邻块的预测模式为L,其上邻块的预测模式为U时,将stat[L][U][C]中的数加1;对所有帧内预测的4×4块重复此操作,即可得到该序列的预测模式统计表。 A preparatory work before coding: the law of large number of values ​​typical sequence prediction mode encoded for statistical experiments to establish the most likely prediction mode under normal circumstances lookup table: Step (A): Intra statistics for each test sequence law prediction mode values: the first prediction mode corresponding to a sequence tables STAT [9] [9] all set to an initial value 0 [9], and then using the 4 × 4 block in the intra prediction encoding individual scans (not including the upper edge of the block and the left block of the image frame edge), each of the first prediction mode specified in the H.264 standard or AVS, the rate-distortion cost value variable setting an initial value of zero, the current prediction mode of the block is C, when the prediction mode of the left neighbor block is L, the prediction mode to which the neighbor block U, the stat [L] [U] number of [C] is incremented by 1; 4 × 4 block is repeated for all the intra prediction this operation, to obtain a prediction mode of the sequence tables.

步骤(B):根据预测模式统计表建立最可能预测模式查找表:用步骤(A)对各测试序列(如AVS标准中规定了112种测试序列)进行统计,每个测试序列得到一张预测模式统计表,将这些表的对应项相加得预测模式统计总表stat_all[9][9][9];分别令左邻块预测模式L和上邻块预测模式U为0至8的所有整数,在每种L和U的组合下,按以下方法确定当前块最可能预测模式:对每个给定的L和U,找出stat_all[L][U][i],i=0、1、……8中的最大值为stat_all[L][U][C],其中C就是该组合下当前块的最可能预测模式,将C写入最可能预测模式查找表的mpm[L][U];如果当前块是图像的上边缘块或左边缘块,在这两种情况下,选择均值(DC)预测模式作为其最可能预测模式,写入最可能预测模式查找表。 Step (B): Statistics establish the most likely prediction mode, the prediction mode look-up table: the step (A) of each test sequence (as specified in the AVS standard 112 test sequences) statistics, to obtain a prediction for each test sequence mode tables, these tables corresponding term-prediction mode obtained by adding the total statistics table stat_all [9] [9] [9]; respectively make a left neighbor block and the prediction mode L o U for all prediction mode of block 0 to 8 integer, in each combination of L and U, by the following method to determine the most probable prediction mode of a current block: for each given L and U, find stat_all [L] [U] [i], i = 0, 1, the maximum value is 8 ...... stat_all [L] [U] [C], where C is the most likely prediction mode is the combination of the current block, writing the most likely prediction mode C lookup table mpm [L] [the U-]; if the current block is an edge block or blocks of the left edge of the image, in both cases, selecting the mean (DC) as the most likely prediction mode is the prediction mode, the prediction mode most likely to write a lookup table.

根据上述方法就可以获得最可能预测模式查找表。 According to the above method can obtain the most likely prediction mode lookup table.

表一为根据AVS测试序列得出的最可能预测模式查找表。 Table 1 gives the look-up table based on the most likely prediction mode AVS test sequences derived.

二.采用下述步骤,在编码端对帧内预测的每个4×4块进行编码。 II. Using the following steps, intra prediction encoding end of each 4 × 4 block is encoded.

(1):根据上邻块和左邻块的预测模式,查最可能预测模式查找表得出当前块的最可能预测模式:如果当前块的左邻块的预测模式为L、上邻块的预测模式U,则当前块的最可能预测模式Cp=mpm[L][U]。 (1): The upper adjacent block and the prediction mode of the left neighbor block, most likely prediction mode search lookup tables stars most probable prediction mode of the current block: if the prediction mode of the current block is a left neighbor block is L, the upper adjacent block prediction mode U, then the most likely prediction mode of a current block Cp = mpm [L] [U].

(2):计算当前块在每一种预测模式下的代价,选择最佳的预测模式:分别在AVS或H.264标准规定的9种可能的预测模式下,设置率失真代价值的初始值,再计算当前块中各象素的原始值和预测值的差值,这些差值的绝对值的和就是该预测模式的代价;选定代价最小的模式为当前块的预测模式C。 (2): the cost of the current block is calculated in each prediction mode, selects the optimum prediction mode: In each of the 9 possible prediction modes specified in the H.264 standard or AVS, the rate-distortion cost value set initial value , and then calculating the difference between the original value of each pixel of the current block and the predicted value, the absolute value of the difference of these is the cost and the prediction mode; minimum consideration mode is selected as the prediction mode of the current block C.

(3):以选定的预测模式对当前块编码:按照AVS标准规定的步骤,首先计算当前块的原始象素值和该预测模式下的预测值之间的差值,得到当前块的残差图像;再对残差图像块进行整数变换和量化;对计算量化后的变换系数进行熵编码;(4):将预测模式信息写入码流:为每个4×4块设定标志位f,如果Cp=C,则令f=1;否则f=0,用预测模式选择变量R表示当前模式信息:如果当前块的预测模式C小于最可能预测模式Cp,则R=C,否则R=C-1,将R对应的三位二进制数写入码流;(5):将熵编码后的信息写入码流;三.再根据下述步骤,在解码端对编码后的图像块进行解码。 (3): the selected prediction mode to encode the current block: step prescribed by the AVS standard, calculates a difference between the first prediction value of the original pixel value and the prediction mode of the current block to obtain a residual of the current block difference image; then the residual image block integer transform and quantization; calculated on the quantized transform coefficient entropy encoding; (4): the prediction mode information written to the code stream: each 4 × 4 block flag is set f, if Cp = C, then let f = 1; otherwise, f = 0, a prediction mode selection variable R indicates the current mode information: if the prediction mode C of the current block is less than the most likely prediction mode Cp, then R = C, or R = C-1, the three-digit binary number R corresponding to the code stream is written; (5): the information is written entropy coding code stream; then according to the following three steps, at the decoding end the encoded image block. decoded.

(1)从码流中读取图像块的帧内编码信息,包括标志位f和模式信息R;(2)确定当前块的最可能预测模式Cp:如果当前块为当前帧图像的左边缘块或上边缘块,则Cp为DC模式;否则设当前块的左邻块的预测模式为L、上邻块的预测模式U,则Cp为预测模式查找表中表项mpm[L][U]的值。 (1) intra-coded image information reading blocks from the code stream, comprising mode information and flag f R; (2) determine the most likely prediction mode of a current block Cp: if the current block to the left edge of the block of the current frame image or the upper edge of the block, the DC mode to Cp; otherwise, set a left neighbor block of a prediction mode of the current block is L, the upper adjacent block prediction modes U, Cp is the prediction mode table lookup table entries mpm [L] [U] value.

(3)确定当前图像块的帧内预测模式C:如果标志位f=1,则C=Cp;否则进行如下处理:如果模式信息R小于Cp,则C=R;否则C=R+1;(4)重建当前图像块:根据预测模式C进行图像块象素的预测;对图像块数据进行熵解码得到变换系数,然后对变换系数进行反量化、反变换;最后得到重建的图像块。 (3) determining a current image block is an intra prediction mode C: If the flag bit f = 1, then C = Cp; otherwise processed as follows: If the mode information R is less than Cp, the C = R; otherwise C = R + 1; (4) reconstructing the current image block: a prediction mode according to the predicted image block pixels C; image block data to obtain transform coefficients entropy decoding, the transform coefficients and inverse quantization, inverse transform; to give the final reconstructed image block.

以中国音视频编码标准化工作组(AVS)制定的面向移动多媒体应用的编码系统AVS-M为例,说明这种基于统计的帧内预测模式编码方法的效果。 In Chinese audio and video coding standardization working group (AVS) developed for mobile multimedia applications AVS-M coding system as an example, we illustrate this coding method based on an intra prediction mode of the statistical results. 在AVS-M编码系统中,采用本发明提出的方法替代原来的方法,对典型序列进行对比实验。 In the AVS-M coding system, using the method of the present invention proposed a method to replace the original, the typical sequence of comparative tests. 表二所示为对比实验的结果。 The results shown in the table as two comparative experiments. 实验选取了CIF 15fs、QCIF 5fs、CIF 30fs、QCIF 15fs四种格式的测试序列。 Experimental test sequence selected CIF 15fs, QCIF 5fs, CIF 30fs, QCIF 15fs four formats. 表中的Y(dB)项表示采用本方法后信噪比改变的分贝数,Bitrate项表示采用本方法后比特率的相对变化量。 Table Y (dB) dB SNR term represents changes in the present method, Bitrate term represents the relative amount of change in the bit rate by this method. 由表中数据可知,每个序列的信噪比都有提高,比特率都有降低。 Seen from the data in the table, each sequence has improved signal to noise ratio, the bit rate has decreased. 这说明本发明提出的基于统计的帧内预测编码方法可在实际应用中更好地提高压缩率和增加预测精度。 This shows that intra prediction encoding method based on statistical can better improve the compression ratio in the practical application of the present invention and the proposed increase in prediction accuracy.

表一 最可能预测模式查找表mpm[L][U] One of the most likely prediction mode table lookup table mpm [L] [U]

表二 对典型序列使用本方法编码后信噪比和比特率的变化 Table II Change after typical sequence for using the present method for encoding the signal to noise ratio and bit rate

Claims (2)

1.一种增强型帧内预测模式编码方法,首先建立最可能预测模式查找表;在编码端根据该查找表对帧内预测的每个块进行编码;在解码端根据该查找表对帧内预测的每个块进行解码;建立最可能预测模式查找表的步骤为:(A1)将测试序列对应的预测模式统计表stat[9][9][9]中设置初始值为0;(A2)对每个图像块进行帧内预测编码处理:首先在中国数字音视频编解码技术标准AVS或国际标准组织ISO和国际电信联盟ITU联合视频工作组制定的视频编码标准H.264规定的每一种预测模式下,设置率失真代价值变量的初始值为零,如果当前块是图像的上边缘块或左边缘块,选择均值DC预测模式作为其最可能预测模式,写入最可能预测模式查找表;否则进行下述处理:(A3)在AVS或H.264标准规定的每一种预测模式下计算其率失真代价值,选定率失真代价值最小的模式C为当前块的 1. An enhanced intra prediction mode encoding method, to establish the most likely prediction mode first lookup table; encoded according to the lookup table for each block in the intra prediction encoding side; the decoding side based on the lookup table frame each block predictive decoding; look-up table step establishing the most likely prediction mode of: (A1) a test sequence corresponding to the prediction mode statistics stat [9] [9] [9] provided the initial value 0; (A2 each first predetermined H.264 video coding standard in China digital audio AVS video coding standard ISO or international standards Organization ITU international Telecommunications Union and developed by joint video Team of:) intra prediction encoding process for each image block the prediction modes, the rate-distortion cost value variable setting an initial value of zero, if the current block is a block edge or the left edge of the image block, selecting a DC prediction mode as the mean of the most likely prediction mode, the write mode is to find the most probable prediction table; otherwise, the following processing is: (A3) in each prediction mode specified in the H.264 standard AVS or calculated rate distortion cost value, the selected minimum rate-distortion cost value of a current block pattern is the C 测模式;在当前块的左邻块的预测模式为L,其上邻块的预测模式为U时,将stat[L][U][C]中的数加1;对所有帧内预测的块重复步骤(A2)至(A3)操作,即得到该序列的预测模式统计表stat[9][9][9];(A4)重复上述步骤(A1)至(A3)对典型测试序列进行统计,每个测试序列得到一张预测模式统计表,将各统计表的对应项相加得预测模式统计总表stat_all[9][9][9];(A5)分别令左邻块预测模式L和上邻块预测模式U为0至8的所有整数,在每种L和U的组合下,按以下方法确定当前块最可能预测模式:对每个给定的L和U,找出stat_all[L][U][i]中的最大值stat_all[L][U][C],其中C就是该组合下当前块的最可能预测模式,i为1到8的任意整数,将C写入最可能预测模式查找表的mpm[L][U];对帧内预测的每个块进行编码的步骤为:(B1)根据上邻块和左邻块的预测模式,查上述最可能预测模式查找表得出当前 Measurement mode; when the prediction mode of the current block is a left neighbor block is L, the upper adjacent block prediction mode is U, the stat [L] [U] Number of [C] is incremented by 1; all the intra-prediction block repeat step (A2) to (A3) operation, i.e., to obtain the prediction mode of the sequence tables stat [9] [9] [9]; (A4) repeating the above steps (A1) to (A3) for a typical test sequence statistics, each test sequence to obtain a prediction mode tables, each corresponding term statistics obtained by adding the total statistical prediction mode table stat_all [9] [9] [9]; (A5), respectively, so that the left adjacent block prediction mode L and upper adjacent block prediction mode for all U an integer from 0 to 8, in each combination of L and U, determining the most probable prediction mode of a current block as follows: for each given L and U, find stat_all [L] [U] [i] is the maximum stat_all [L] [U] [C], where C is the most likely prediction mode of the combination of the current block, i is an arbitrary integer from 1 to 8, will be written in C the most likely prediction mode table lookup mpm [L] [U]; step for each block of intra prediction encoding of: (B1) the upper adjacent block and the prediction mode of the left neighbor block is most likely to check the above-described prediction lookup table mode current draw 的最可能预测模式:设当前块的左邻块的预测模式为L、上邻块的预测模式U,如果当前图像块为上边缘块或左边缘块,则Cp=DC;否则当前块的最可能预测模式Cp=mpm[L][U];(B2)计算当前块在每一种预测模式下的代价值,选择最佳的预测模式:分别在AVS标准或H.264标准规定的9种可能的预测模式下,设置率失真代价值的初始值,然后计算当前块的率失真代价值;选定率失真代价值最小的模式为当前块的预测模式C;(B3)以选定的预测模式C对当前块编码:首先计算当前块的原始象素值和该预测模式C下的预测值之间的差值,得到当前块的残差图像;再对残差图像块进行整数变换和量化;对计算量化后的变换系数进行熵编码;(B4)将预测模式信息写入码流:为每个4×4块设定标志位f,如果Cp=C,则令f=1,否则f=0,将标志位f用1比特写入码流;用变量R表示当前要 The most likely prediction mode: Set the current block is prediction mode of a left neighbor block is L, the prediction mode of the neighbor block U, if the current image block to the upper edge of the block or a left edge of the block, then the Cp = DC; otherwise, most of the current block may Cp = mpm [L] [U] prediction mode; (B2) to blocks of the current value is calculated in each prediction mode, selects the optimum prediction mode: AVS are specified in the standard or H.264 standard nine kinds the possible prediction modes, the rate-distortion cost value set initial value, and then calculates rate-distortion cost value of a current block; substituting the minimum rate-distortion value of the selected mode is the prediction mode of the current block C; (B3) to the selected prediction encoding the current block pattern C: firstly calculating the difference between the original pixel value in the predicted value of the current block and the prediction mode C, giving a residual image of the current block; and then the residual image block integer transform and quantization ; calculated on the quantized transform coefficient entropy encoding; (B4) prediction mode information of the code stream is written: for each 4 × 4 block setting flag bits f, if Cp = C, then let f = 1, otherwise f = 0, the flag bit f is written with a bit stream; represented by a variable current to R 入的模式信息:如果当前块的预测模式C小于最可能预测模式Cp,则R=C,否则R=C-1,将R对应的三位二进制数写入码流;(B5)将熵编码后的信息写入码流;对帧内预测的每个4×4块进行解码的步骤为:(C1)从码流中读取图像块的帧内编码信息,包括标志位f和模式信息R;(C2)确定当前块的最可能预测模式Cp:如果当前块为当前帧图像的左边缘块或上边缘块,则Cp为DC模式;否则设当前块的左邻块的预测模式为L、上邻块的预测模式U,则Cp为预测模式查找表中表项mpm[L][U]的值;(C3)确定当前图像块的帧内预测模式C:如果标志位f=1,则C=Cp;否则进行如下处理:如果模式信息R小于Cp,则C=R,否则C=R+1;(C4)重建当前图像块:根据预测模式C进行图像块象素的预测;对图像块数据进行熵解码得到变换系数,然后对变换系数进行反量化、反变换;最后得到重建的图像 The mode information: if the current prediction mode of the block C is less than the prediction mode most likely Cp, then R = C, or R = C-1, the three-digit binary number R corresponding to the code stream is written; (B5) The entropy encoding after the information is written in the code stream; each 4 × 4 block decoding step of intra prediction is: (C1) an image information reading intra-coded blocks from the code stream, comprising mode information and flag f R ; (C2) to determine the most likely prediction mode Cp current block: if the current block to the left edge of the block of the current frame image or the upper edge of the block, the Cp is a DC mode; otherwise, set the current left prediction mode of the block neighboring block is L, prediction modes neighbor block U, the lookup value Cp mpm [L] [U] is a table of the prediction mode table entry; (a C3) determines the current block is an intra prediction mode C: If the flag bit f = 1, then C = Cp; otherwise processed as follows: If the mode information R is less than Cp, the C = R, or C = R + 1; (C4) reconstructing the current image block: prediction image pixel block according to the prediction mode C; image entropy decoding the block data to obtain transform coefficients, the transform coefficients and inverse quantization, inverse transform; finally obtained image reconstruction 块。 Piece.
2.根据权利要求1所述的方法,其特征在于:在面向移动通信的中国音视频编解码标准AVS-M中,最可能预测模式查找表mpm[L][U](L,U=0,1,...8)为: The method according to claim 1, wherein: audio and video codec in China AVS-M standard for mobile communications, the most likely prediction mode table lookup mpm [L] [U] (L, U = 0 , 1, ... 8) as follows: .
CN 200510018489 2005-03-31 2005-03-31 An enhanced in-frame predictive mode coding method CN100348051C (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200510018489 CN100348051C (en) 2005-03-31 2005-03-31 An enhanced in-frame predictive mode coding method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200510018489 CN100348051C (en) 2005-03-31 2005-03-31 An enhanced in-frame predictive mode coding method

Publications (2)

Publication Number Publication Date
CN1674680A CN1674680A (en) 2005-09-28
CN100348051C true CN100348051C (en) 2007-11-07

Family

ID=35046896

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200510018489 CN100348051C (en) 2005-03-31 2005-03-31 An enhanced in-frame predictive mode coding method

Country Status (1)

Country Link
CN (1) CN100348051C (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107770535A (en) * 2011-11-04 2018-03-06 杰尼普Pte有限公司 The method for producing reconstructed blocks

Families Citing this family (39)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100442857C (en) 2005-10-12 2008-12-10 华为技术有限公司 Method of enhanced layer in-frame predicting method and encoding and decoding apparatus
JP5089878B2 (en) 2005-10-28 2012-12-05 パナソニック株式会社 Image encoding device
WO2007053126A1 (en) 2005-11-07 2007-05-10 Agency For Science, Technology And Research Methods and devices for decoding and encoding data
CA2656123A1 (en) * 2006-04-28 2007-11-08 Ntt Docomo, Inc. Image predictive coding device, image predictive coding method, image predictive coding program, image predictive decoding device, image predictive decoding method and image predictive decoding program
JP2007300380A (en) * 2006-04-28 2007-11-15 Ntt Docomo Inc Image predictive encoding device, image predictive encoding method, image predictive encoding program, image predictive decoding device, image predictive decoding method, and image predictive decoding program
CN100596203C (en) 2006-06-05 2010-03-24 华为技术有限公司;华中科技大学 Image coding/decoding method and system
CN100508610C (en) 2007-02-02 2009-07-01 清华大学 Method for quick estimating rate and distortion in H.264/AVC video coding
CN100568952C (en) 2007-03-29 2009-12-09 上海大学;上海广电(集团)有限公司中央研究院 Real time fine flexible coding method based on H.264
EP2156672A1 (en) * 2007-05-04 2010-02-24 QUALCOMM Incorporated Video coding mode selection using estimated coding costs
CN100566427C (en) * 2007-07-31 2009-12-02 北京大学;北京北大方正电子有限公司 Intraframe prediction coding optimum mode selecting method and apparatus for video coding
CN101415121B (en) 2007-10-15 2010-09-29 华为技术有限公司 Self-adapting method and apparatus for forecasting frame
CN101500161B (en) 2008-01-31 2012-03-21 华为技术有限公司 Inter-frame prediction method and apparatus based on adaptive block transformation
KR20090095316A (en) * 2008-03-05 2009-09-09 삼성전자주식회사 Method and apparatus for image intra prediction
CN101605255B (en) 2008-06-12 2011-05-04 华为技术有限公司 Method and device for encoding and decoding video
CN101677406B (en) 2008-09-19 2011-04-20 华为技术有限公司 Method and apparatus for video encoding and decoding
CN101729883B (en) * 2008-10-14 2013-03-20 上海炜寅信息科技有限公司 Method for adaptively adjusting video decoding complexity
CN101409845B (en) 2008-10-31 2010-10-06 北京大学软件与微电子学院 Method and apparatus for estimating video distortion in AVS video encoding
CN101945270B (en) * 2009-07-06 2013-06-19 联发科技(新加坡)私人有限公司 Video coder, method for internal prediction and video data compression
WO2012008515A1 (en) * 2010-07-15 2012-01-19 シャープ株式会社 Decoding device and coding device
ES2602779T3 (en) * 2010-08-17 2017-02-22 M&K Holdings Inc. Intra-prediction decoding procedure
US9204155B2 (en) * 2010-09-30 2015-12-01 Futurewei Technologies, Inc. Multiple predictor set for intra coding with intra mode prediction
CN102724493B (en) * 2011-01-18 2014-06-25 清华大学 Coding and decoding methods of intra-frame prediction modes based on image blocks and codec
WO2012113197A1 (en) * 2011-02-24 2012-08-30 中兴通讯股份有限公司 Encoding or decoding method of prediction mode, and device thereof
CN102685474B (en) * 2011-03-10 2014-11-05 华为技术有限公司 Encoding and decoding method of prediction modes, encoding and decoding device and network system
CN105187834B (en) * 2011-06-17 2017-06-27 Jvc建伍株式会社 Picture decoding apparatus, picture decoding method and reception device, method of reseptance
DK2887670T3 (en) * 2011-06-28 2017-09-25 Samsung Electronics Co Ltd Method and apparatus for decoding video method and apparatus for decoding video, followed by intra-prediction.
CN104918055B (en) 2011-06-28 2019-07-26 三星电子株式会社 Method and apparatus for using intra prediction to carry out image coding and decoding
CN102857750B (en) * 2011-07-01 2015-05-06 华为技术有限公司 Method, device and system for in-frame predictive coding and encoding
JP5728671B2 (en) * 2011-10-24 2015-06-03 インフォブリッジ ピーティーイー. エルティーディー. Video decoding device
CN107105262A (en) * 2011-10-24 2017-08-29 英孚布瑞智有限私人贸易公司 The method decoded to intra prediction mode
CN102364948B (en) * 2011-10-28 2013-10-16 上海国茂数字技术有限公司 Method for two-way compensation of video coding in merging mode
KR20130049523A (en) * 2011-11-04 2013-05-14 오수미 Apparatus for generating intra prediction block
KR20130049522A (en) * 2011-11-04 2013-05-14 오수미 Method for generating intra prediction block
KR20130049524A (en) * 2011-11-04 2013-05-14 오수미 Method for generating intra prediction block
US9554136B2 (en) 2011-12-05 2017-01-24 Lg Electronics Inc. Method and device for intra prediction
US9210438B2 (en) * 2012-01-20 2015-12-08 Sony Corporation Logical intra mode naming in HEVC video coding
CN104168481B (en) * 2014-07-24 2017-08-04 北京大学 For the soft coding methods of real-time AVS of Intel mobile platforms
CN104702959B (en) * 2015-03-20 2019-03-26 上海国茂数字技术有限公司 A kind of intra-frame prediction method and system of Video coding
CN107222743A (en) * 2017-07-04 2017-09-29 苏睿 A kind of image processing method, device and system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107770535A (en) * 2011-11-04 2018-03-06 杰尼普Pte有限公司 The method for producing reconstructed blocks

Also Published As

Publication number Publication date
CN1674680A (en) 2005-09-28

Similar Documents

Publication Publication Date Title
JP4916579B2 (en) Interlaced video encoding and decoding
CN1232122C (en) Method and apparatus for coding and or decoding moving image
CN100373950C (en) Adaptive de-blocking filtering apparatus and method for MPEG video decoder
JP4480713B2 (en) Method and apparatus for improved coding mode selection
JP5530181B2 (en) Image encoding method and image decoding method
KR101574866B1 (en) Performing motion vector prediction for video coding
RU2377737C2 (en) Method and apparatus for encoder assisted frame rate up conversion (ea-fruc) for video compression
JP5401009B2 (en) Video intra prediction encoding and decoding method and apparatus
CN101080014B (en) Video decoding method and apparatus
US7327786B2 (en) Method for improving rate-distortion performance of a video compression system through parallel coefficient cancellation in the transform
JP4927753B2 (en) Moving picture coding method, moving picture decoding method, moving picture coding apparatus, moving picture decoding apparatus, and semiconductor device
KR101622450B1 (en) Video encoding and decoding using transforms
EP1457056B1 (en) Skip macroblock coding
US7266149B2 (en) Sub-block transform coding of prediction residuals
KR20090005080A (en) Method and system for coding and decoding information associated with video compression
CN1224268C (en) Moving picture coding apparatus and moving picture decoding apparatus
CN101167364B (en) Scalable video coding with two layer encoding and single layer decoding
CN100579229C (en) Method for processing interlaced video with a plurality of reference diagrams
US20050013500A1 (en) Intelligent differential quantization of video coding
KR20170016994A (en) Unified merge mode and adaptive motion vector prediction mode candidates selection
US20070217500A1 (en) Encoding Method for Skipped Macroblocks in Video Image
TWI392374B (en) Method and apparatus for using frame rate up conversion techniques in scalable video coding
US8213503B2 (en) Skip modes for inter-layer residual video coding and decoding
US20080025390A1 (en) Adaptive video frame interpolation
US20090196350A1 (en) Methods and devices of intra prediction encoding and decoding

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C14 Granted
C17 Cessation of patent right