CN1137211A - Improved motion compensation method for use in image encoding system - Google Patents

Abstract

Translated fromChinese

一种能够改善图象质量的方法，包括下述步骤：(a)利用一种块匹配算法为当前帧中的各搜索块检测出一个运动矢量；(b)将一个搜索块分成一个位于该搜索块中央的中心区及一个位于该中心区外侧的边沿区；(c)确定该搜索块的运动矢量作为中心区的一个最佳运动矢量；以及(d)根据该搜索块及某一个或多个相邻的搜索块的运动矢量，为该搜索块的边沿区确定多个最佳运动矢量。

A method capable of improving image quality, comprising the steps of: (a) detecting a motion vector for each search block in a current frame using a block matching algorithm; (b) dividing a search block into a block located in the search block A central area in the center of the block and an edge area located outside the central area; (c) determine the motion vector of the search block as an optimal motion vector in the central area; and (d) according to the search block and one or more For the motion vectors of the adjacent search blocks, a plurality of optimal motion vectors are determined for the border area of the search block.

Description

Translated fromChinese

用在图象编码系统中的改进的运动补偿方法An Improved Motion Compensation Method Used in Image Coding System

本发明涉及用在图象编码系统中的运动补偿方法；更具体地，涉及一种能够消除出现在解码图象信号中的分块现象的改进的方法。The present invention relates to a motion compensation method used in an image coding system; more particularly, to an improved method capable of eliminating blocking phenomena occurring in a decoded image signal.

众所周知，传输数字化视频信号能提供比传输模拟信号质量高得多的视频图象。当包含一序列图象“帧”的一个图象信号以数字形式表示时，便生成大量用于传输的数据，尤其是在高清晰度电视系统的情况中。然而，由于一条传统的传输信道的可利用的频带宽度是有限的，为了通过有限的信道带宽传输大量的数字数据，不可避免地要压缩或减少传输的数据量。在各种视频压缩技术中，将时间和空间压缩技术与统计编码技术组合在一起的所谓混合编码技术已知是最有效的。It is well known that the transmission of digitized video signals can provide much higher quality video images than the transmission of analog signals. When an image signal comprising a sequence of image "frames" is represented in digital form, large amounts of data are generated for transmission, especially in the case of high definition television systems. However, since the available frequency bandwidth of a traditional transmission channel is limited, in order to transmit a large amount of digital data through the limited channel bandwidth, it is inevitable to compress or reduce the amount of transmitted data. Among various video compression techniques, a so-called hybrid coding technique that combines temporal and spatial compression techniques with statistical coding techniques is known to be the most effective.

大多数混合编码技术采用运动补偿的DPCM9(差分脉冲码调制)、二维DCT(离散余弦变换)、DCT系数的量化、及VLC(可变长度编码)。运动补偿的DPCM为确定一个目标在一个当前帧及其前一帧之间的运动，并根据该目标的运动流预测当前帧以生成表示当前帧与其预测之间的差的一个差分信号的过程。Most hybrid coding techniques employ motion compensated DPCM9 (Differential Pulse Code Modulation), two-dimensional DCT (Discrete Cosine Transform), quantization of DCT coefficients, and VLC (Variable Length Coding). Motion-compensated DPCM is the process of determining the motion of an object between a current frame and its previous frame, and predicting the current frame from the motion flow of the object to generate a differential signal representing the difference between the current frame and its prediction.

减少或者除诸如运动补偿的DPCM数据等图象数据之间的空间冗余性的二维DCT将一块数字图象数据(例如，8×8个象素的一块)转换成一组变换系数数据。Two-dimensional DCT, which reduces or eliminates spatial redundancy between image data such as motion compensated DPCM data, converts a block of digital image data (eg, a block of 8x8 pixels) into a set of transform coefficient data.

具体地说，在运动补偿的DPCM中，当前帧数据是根据当前与前面的帧之间的一种运动估计从前一帧数据中预测的。这种估计的运动可以用表示前面的与当前的帧之间的象素位移的二维运动矢量来描述。Specifically, in motion-compensated DPCM, current frame data is predicted from previous frame data based on a motion estimate between the current and previous frames. This estimated motion can be described by a two-dimensional motion vector representing the pixel displacement between the previous and current frames.

曾经提出过若干种用于估计一个目标在视频序列中的位移的方法。通常，可将它们分成两类：象素递归算法及块匹配算法。本发明主要考虑块匹配算法。Several methods have been proposed for estimating the displacement of an object in a video sequence. In general, they can be divided into two categories: pixel recursive algorithms and block matching algorithms. The present invention primarily considers block matching algorithms.

按照块匹配算法，将一个当前帧分成多个搜索块，。一个搜索块的大小通常在8×8与32×32个象素之间。为了确定当前帧中的一个搜索块的运动矢量，在当前帧的搜索块与包含在前一帧内的一个通常较大的搜索区中的多个相同大小的候选块的每一个之间执行相似性计算。采用诸如平均绝对误差或均方误差等误差函数来进行当前帧的搜索块与搜索区中的各候选块之间的相似性测定。根据定义，运动矢量代表搜索块与得出最小误差函数的一个候选块之间的位移。Divide a current frame into multiple search blocks according to the block matching algorithm. The size of a search block is usually between 8x8 and 32x32 pixels. To determine the motion vector for a search block in the current frame, a similarity is performed between the search block in the current frame and each of a plurality of candidate blocks of the same size contained in a usually larger search area in the previous frame. sexual calculation. The similarity measurement between the search block of the current frame and each candidate block in the search area is performed using an error function such as mean absolute error or mean square error. By definition, a motion vector represents the displacement between the search block and a candidate block that yields the smallest error function.

经过编码的图象数据是通过一条传统的传输信道传输到包含在一个图象信号解码系统中的一个图象信号解码器的，它执行编码操作的逆过程，借此重构原来的图象数据。然而，重构的图象数据通常呈现称作分块现象的令人不愉快的人为现象，其中一块的边界线在接收端上成为看得见的。由于一帧是以块为单位编码的，所以出现这种分块后果。The encoded image data is transmitted through a conventional transmission channel to an image signal decoder included in an image signal decoding system, which performs the inverse process of the encoding operation, thereby reconstructing the original image data . However, reconstructed image data often exhibits an unpleasant artifact known as blocking, in which the boundary lines of a block become visible at the receiving end. This blocking effect occurs because a frame is coded in blocks.

因此，本发明的一个主要目的为提供一种用在图象编码系统中的运动补偿方法，该方法能够消除出现在图象信号的一块的边界上的分块后果，借此改善该系统所生成的图象的质量。Therefore, a main object of the present invention is to provide a motion compensation method used in an image coding system, which can eliminate the blocking effect occurring on the boundary of a block of the image signal, thereby improving the performance generated by the system. image quality.

按照本发明，提供了一种确定视频信号的一个当前帧与前一帧之间的最佳运动矢量的方法，其中将该当前帧分成多个大小相等的搜索块，而前一帧中则包含对应数目的搜索区，各搜索区具有多个所述相等大小的候选块，该方法包括下述步骤：According to the present invention, there is provided a method of determining the best motion vector between a current frame of a video signal and a previous frame, wherein the current frame is divided into a plurality of search blocks of equal size, while the previous frame contains A corresponding number of search areas, each search area has a plurality of candidate blocks of equal size, the method comprises the following steps:

(a)采用一种块匹配算法为当前帧中的各搜索块检测出一个运动矢量；(a) using a block matching algorithm to detect a motion vector for each search block in the current frame;

(b)将一个搜索块分成位于该搜索块中央的一个中心区及位于该中心区外侧的一个边沿区；(b) dividing a search block into a central area located in the center of the search block and a border area located outside the central area;

(c)确定该搜索块的运动矢量作为中心区的一个最佳运动矢量，(c) determining the motion vector of the search block as an optimal motion vector for the central area,

(d)根据该搜索块及其一个或多个邻接搜索块的运动矢量，为该搜索块的边沿区确定多个最佳运动矢量。(d) Determining a plurality of optimal motion vectors for the border region of the search block based on the motion vectors of the search block and one or more adjacent search blocks.

从下面结合附图给出的较佳实施例的描述中，本发明的上述与其它目的与特征将是显而易见的，附图中：From the following description of the preferred embodiment given in conjunction with the accompanying drawings, the above-mentioned and other objects and features of the present invention will be apparent, in the accompanying drawings:

图1是提供采用本发明的运动补偿装置的一个图象信号编码系统的方框图；Fig. 1 is a block diagram providing an image signal encoding system employing the motion compensation device of the present invention;

图2示出图1的运动补偿装置的详细方框图；以及Figure 2 shows a detailed block diagram of the motion compensation device of Figure 1; and

图3示出说明在本发明的运动补偿装置中所进行的边沿区形成的图。FIG. 3 is a diagram illustrating formation of a margin region performed in the motion compensation device of the present invention.

参见图1，其中示出用于压缩一个数字视频信号的编码装置10的方框图，它包括本发明的运动补偿装置150，其中该数字视频信号包含视频信号的多个帧。Referring to FIG. 1, there is shown a block diagram of anencoding apparatus 10 for compressing a digital video signal comprising a plurality of frames of video signal, which includes themotion compensation apparatus 150 of the present invention.

编码装置10包括一个第一帧存储器100、一个减法器102、一个图象信号编码器105、一个熵编码器107、一个图象信号解码器113、一个加法器115、一个第二帧存储器124以及该运动补偿装置150。Encoding device 10 comprises afirst frame memory 100, asubtractor 102, animage signal encoder 105, anentropy encoder 107, animage signal decoder 113, anadder 115, asecond frame memory 124 and Themotion compensation device 150 .

将包含在一个输入视频信号中的一个当前帧存储在第一帧存储器100中，该存储器通过线路L9连接在减法器102上并通过线路L10连接在运动补偿装置150上。所存储的当前帧是在逐块地基础上读取的，其中块大小通常在8×8与32×32个象素的范围内。A current frame contained in an input video signal is stored in afirst frame memory 100 which is connected via line L9 tosubtractor 102 and via line L10 to motion compensation means 150 . The stored current frame is read on a block-by-block basis, where block sizes are typically in the range of 8x8 and 32x32 pixels.

本发明的运动补偿装置150首先通过采用传统的块匹配算法为当前帧的各搜索块检测一个运动矢量，该运动矢量表示该当前帧中的各搜索块与包含在来自第二帧存储器124的前一帧的一个对应搜索区中的与之最相似的候选块之间的空间位移；以及根据该搜索块及其相邻的搜索块的运动矢量，为包含在各搜索块中的象素确定多个最佳运动矢量。此后，运动补偿装置150利用所确定的最佳运动矢量从第二帧存储器124中检索前一帧的对应象素值，借此提供一个预测的当前帧。将搜索块的运动矢量与预测的当前帧信号分别馈送给熵编码器107、减法器102与加法器113。下面参照图2与3描述运动补偿装置150的细节。Themotion compensation device 150 of the present invention firstly detects a motion vector for each search block of the current frame by using a conventional block matching algorithm, and the motion vector represents the relationship between each search block in the current frame and the previous frame from thesecond frame memory 124. The spatial displacement between the most similar candidate blocks in a corresponding search area of a frame; and according to the motion vectors of the search block and its adjacent search blocks, determine how many best motion vector. Thereafter, the motion compensation means 150 uses the determined optimal motion vector to retrieve the corresponding pixel values of the previous frame from thesecond frame memory 124, thereby providing a predicted current frame. The motion vector of the search block and the predicted current frame signal are respectively fed to theentropy encoder 107 , thesubtractor 102 and theadder 113 . Details of themotion compensation device 150 are described below with reference to FIGS. 2 and 3 .

在减法器102上从通过线路L9提供的当前帧信号中减去来自运动补偿装置150的预测的当前帧信号，并将表示当前帧的与预测的当前帧的象素值之间的差的得出的数据，即误差信号，发送给图象信号编码器105，在其中用诸如DCT与任何已知的量化方法将该误差信号码编成多组量化的变换系数。The predicted current frame signal from themotion compensation device 150 is subtracted from the current frame signal provided by the line L9 at thesubtractor 102, and the result of the difference between the pixel values representing the current frame and the predicted current frame is obtained The output data, that is, the error signal, is sent to theimage signal encoder 105, where the error signal is coded into sets of quantized transform coefficients using, for example, DCT and any known quantization method.

此后，将量化的变换系数传输给熵编码器107与图象信号解码器113。图象信号解码器113利用逆量化及逆离散余弦变换将来自图象信号编码器105的量化的变换系数转换回一个重构的误差信号，在加法器115上，将从图象信号解码器113接收的重构的误差信号及通过线路L30从运动补偿装置150提供的预测的当前帧信号进行组合以提供一个重构的当前帧信号，供作为一个前一帧信号存储在第二帧存储器124中。Thereafter, the quantized transform coefficients are transmitted to theentropy encoder 107 and theimage signal decoder 113 . Theimage signal decoder 113 utilizes inverse quantization and inverse discrete cosine transform to convert the quantized transform coefficient from theimage signal encoder 105 back into a reconstructed error signal. The received reconstructed error signal and the predicted current frame signal provided via line L30 from the motion compensation means 150 are combined to provide a reconstructed current frame signal for storage in thesecond frame memory 124 as a previous frame signal .

在熵编码器107上，用诸如可变长度编码技术将图象信号编码器105供给的量化的变换系数及通过线路L20从运动补偿装置150发送来的运动矢量编码在一起。此后，将编码信号提供给一台发送机(未示出)供其发送。In theentropy encoder 107, the quantized transform coefficient supplied from theimage signal encoder 105 and the motion vector sent from themotion compensation device 150 through the line L20 are coded together by a variable length coding technique, for example. Thereafter, the encoded signal is provided to a transmitter (not shown) for transmission.

现在参见图2，其中示出了图1中所示的运动补偿装置150的细节。该运动补偿装置150包括一个运动估计器209、一个存储器210、一个区形成单元213、一个最佳运动矢量确定单元216及一个运动补偿器219。Referring now to FIG. 2 , details of themotion compensation device 150 shown in FIG. 1 are shown. Themotion compensation device 150 includes amotion estimator 209 , amemory 210 , aregion forming unit 213 , an optimal motionvector determining unit 216 and amotion compensator 219 .

首先，运动估计器209从第一帧存储器100中检索当前帧信号并从第二帧存储器124中检索前一帧信号，并通过利用本技术中众所周知的传统块匹配技术检测一个表示当前帧中的各搜索块与前一帧中与之最相似的一个块之间的空间位移的运动矢量。将来自运动估计器209的各搜索区的运动矢量提供给图1中所示的熵编码器107及存储器210，在其中存储当前帧的搜索块的运动矢量。First, themotion estimator 209 retrieves the current frame signal from thefirst frame memory 100 and the previous frame signal from thesecond frame memory 124, and detects a signal representing the current frame by using conventional block matching techniques well known in the art. The motion vector of the spatial displacement between each search block and the most similar block in the previous frame. The motion vectors of the respective search areas from themotion estimator 209 are supplied to theentropy encoder 107 shown in FIG. 1 and thememory 210 in which the motion vectors of the search blocks of the current frame are stored.

同时，将当前帧信号提供给区形成单元213，在其中将各搜索块分成一个边沿区及一个中心区。在图3中，示出了在按照本发明的较佳实施例的区形成单元213上进行的一种区形成方案。如图3中所示，将诸如16×16个象素的SB5等各搜索块分成诸如12×12个象素的位于搜索块SB5中央的一个中心区CR5，及由位于该中心区CR5外侧的象素构成的一个边沿区，其中该边沿区包含各由2×12个象素构成的四个边区ER5-1至ER5-4及各由2×2个象素构成的四个角区CR5-1至CR5-4。将表示当前帧中的中心、边与角区的位置的区信息提供给最佳运动矢量确定单元216。At the same time, the current frame signal is supplied to thearea forming unit 213, in which each search block is divided into a border area and a central area. In FIG. 3, a region forming scheme performed on aregion forming unit 213 according to a preferred embodiment of the present invention is shown. As shown in FIG. 3, each search block such as SB5 of 16*16 pixels is divided into a central area CR5 located in the center of the search block SB5 of, for example, 12*12 pixels, and a central area CR5 located outside the central area CR5. A border region composed of pixels, wherein the border region includes four border regions ER5-1 to ER5-4 each consisting of 2×12 pixels and four corner regions CR5-4 each consisting of 2×2 pixels 1 to CR5-4. The region information representing the positions of the center, side and corner regions in the current frame is supplied to the optimum motionvector determination unit 216 .

最佳运动矢量确定单元216从存储器210中检索当前帧中的搜索块的运动矢量，并响应区形成单元213所提供的区信息，利用各搜索块与其相邻的搜索块的运动矢量为各搜索块的区确定最佳运动矢量。具体地，通过求出诸如SB5与SB2这两个搜索块的运动矢量的平均值而得出位于沿两个搜索块，即SB5与SB2，之间的一条边界上的一个边区ER5-1的一个最佳运动矢量。类似地，通过分别求出两个搜索块SB5与SB4、SB5与SB6及SB5与SB8的运动矢量的平均值而确定边区ER5-2至ER5-4的最佳运动矢量。另一方面，通过求出包含一个角区的一个搜索块的运动矢量与在该角区上与该搜索块相连的搜索块的矢量的平均值而得出该角区的最佳运动矢量。例如求出搜索块SB1、SB2、SB4与SB5的运动矢量的平均值而得出角区CR5-1的最佳运动矢量。同样，角区CR5-2至CR5-4的最佳运动矢量是根据搜索块SB2至SB9的运动矢量确定的。至于诸如CR5等中心区的最佳运动矢量，是指定包含该中心区CR5的搜索块SB5的运动矢量作为其最佳运动矢量的。The optimal motionvector determining unit 216 retrieves the motion vectors of the search blocks in the current frame from thememory 210, and responds to the area information provided by thearea forming unit 213, and uses the motion vectors of each search block and its adjacent search blocks for each search block. The region of the block determines the best motion vector. Specifically, an edge region ER5-1 located along a border between two search blocks, SB5 and SB2, is obtained by averaging the motion vectors of the two search blocks such as SB5 and SB2. Best motion vector. Similarly, the optimum motion vectors of the border regions ER5-2 to ER5-4 are determined by averaging the motion vectors of the two search blocks SB5 and SB4, SB5 and SB6, and SB5 and SB8, respectively. On the other hand, an optimum motion vector for a corner area is obtained by obtaining an average value of a motion vector of a search block including a corner area and a vector of a search block connected to the search block on the corner area. For example, the average value of the motion vectors of the search blocks SB1, SB2, SB4 and SB5 is obtained to obtain the optimal motion vector of the corner region CR5-1. Likewise, the optimum motion vectors of the corner regions CR5-2 to CR5-4 are determined based on the motion vectors of the search blocks SB2 to SB9. As for the optimum motion vector of a central area such as CR5, the motion vector of the search block SB5 containing this central area CR5 is designated as its optimum motion vector.

以这一方式并按照本发明，诸如CR1-1、CR2-2、CR4-1与CR5-1等互相连接的任何角区具有通过求出包含这些角区的搜索块(诸如SB1、SB2、SB4与SB5)的运动矢量的平均值而确定的相同的最佳运动矢量，而诸如ER2-1与ER5-1等任何相连的边区则共用一个通过求出包含这些边区的搜索块(诸如SB2与SB5)的运动矢量的平均值而计算出的最佳运动矢量。In this way and according to the present invention, any corner regions interconnected such as CR1-1, CR2-2, CR4-1, and CR5-1 have The same optimal motion vector is determined by the average of the motion vectors of SB5), while any connected border regions such as ER2-1 and ER5-1 share a search block (such as SB2 and SB5) that contains these border regions. ) The best motion vector calculated by the average of the motion vectors.

然后将当前帧的各搜索块的最佳运动矢量提供给一个运动补偿器219。The best motion vectors of the search blocks of the current frame are then supplied to amotion compensator 219 .

运动补偿器219利用最佳运动矢量确定单元216所提供的最佳运动矢量从图1中所示的第二帧存储器124中检索各象素值，借此通过线路L30将预测的当前帧信号提供给图1中所示的减法器102与加法器115。Motion compensator 219 utilizes the optimal motion vector provided by optimal motionvector determination unit 216 to retrieve each pixel value fromsecond frame memory 124 shown in Fig. 1, thereby provides the current frame signal of prediction by line L30 to thesubtractor 102 andadder 115 shown in FIG. 1 .

在一个与本发明的编码装置10对应的数字视频信号解码装置中，根据从编码装置10传输来的运动矢量，以相对于区形成单元213及最佳运动矢量确定单元描述的类似方式，重构最佳运动矢量，借此根据所传输的表示误差信号与运动矢量的编码信号便能重构当前帧信号。In a digital video signal decoding device corresponding to theencoding device 10 of the present invention, based on the motion vector transmitted from theencoding device 10, in a manner similar to that described with respect to theregion forming unit 213 and the optimal motion vector determining unit, reconstruction An optimal motion vector is used, whereby the current frame signal can be reconstructed according to the transmitted coded signal representing the error signal and the motion vector.

虽然已经相对于特定的实施例描述了本发明，熟悉本技术的人员显而易见可以作出各种改变与修正，而不脱离下述权利要求书中所定义的发明范围。Although the invention has been described with respect to particular embodiments, it will be apparent to those skilled in the art that various changes and modifications can be made without departing from the scope of the invention as defined in the following claims.

Claims (5)

1, a kind of method that is used for determining the present frame and the optimum movement vector between the former frame of vision signal, wherein this present frame is divided into equal-sized a plurality of search block, the field of search that then comprises corresponding number in the former frame, each field of search has a plurality of candidate blocks of described equal sizes, and this method comprises the steps:

(a) utilize a kind of block matching algorithm to detect a motion vector for each search block in the present frame;

(b) search block is divided into a center and an edge district that is positioned at this outside, center that is positioned at these search block central authorities;

(c) motion vector of determining this search block is as the good speed fortune vector of this center; And

(d) according to this search block and one or more motion vector thereof, for the edge district of this search block determining a plurality of optimum movement vectors in abutting connection with search block.

2, according to the process of claim 1 wherein the mean value of the motion vector of optimum movement vector by obtaining this search block and one or more adjacent search block thereof in this edge district and definite.

3, according to the method for claim 1, wherein the edge district of this search block comprises four border areas and four angular regions, wherein each angular region is positioned on the angle of this search block, each border area is then along on the limit of this search block between two angular regions, described each angular region has three adjacent search block, and described border area then has an adjacent search block.

4, according to the method for claim 3, an optimum movement vector of described each angular region is to determine according to four motion vectors of its three adjacent search block and this search block, and an optimum movement vector in described each border area is to determine according to the adjacent search block of one and two motion vectors of this search block.

5, according to the method for claim 4, the optimum movement vector of described each angular region is definite by the mean value of obtaining described four motion vectors, and the optimum movement vector in described each border area then is definite by the mean value of obtaining described two motion vectors.

Images