CN100385957C - A Method of Motion Vector Prediction - Google Patents

Abstract

The present invention relates to the technical field of video image treatment, particularly to a method for predicting motion vectors. Distances between motion vectors of adjacent macro blocks or blocks are used for serving as a judgment criterion to select the motion vectors of the current macro blocks or the blocks. The present invention has the following steps: (1) the distances between the motion vectors of the adjacent macro blocks or the blocks are separately calculated; (2) the distances between the motion vectors are sorted and compared, and the motion vectors of the blocks or the macro blocks, which are corresponding to middle distances, are selected to serve as predicted values of the motion vectors of the current blocks or the macro blocks. The method is suitable for video coding of any inter prediction.

Description

A Method of Motion Vector Prediction

technical field

The invention relates to the technical field of video image processing, in particular to a method for motion vector prediction.

Background technique

With the rapid development and wide application of multimedia technology, various video data are used more and more widely. Since the original video data requires a huge bandwidth and has a lot of redundancy, it is usually stored or transmitted after being compressed by encoding. Among them, inter-frame prediction is the key technology to remove temporal redundancy. If the motion vector field of inter-frame prediction is consistent with the actual motion vector field, then the compression efficiency of video coding will be greatly increased. Otherwise, inter-frame redundancy may not be able to It is effectively removed, thus affecting the coding efficiency of the entire system. Motion vector prediction is the starting point for calculating the motion vector field. If the motion vector field can be effectively predicted, on the one hand, the complexity and calculation amount of motion estimation will be reduced. On the other hand, it will also help to remove inter-frame redundancy and improve the overall system coding efficiency.

Video coding began in the late 1980s. From MPEG-1, H.261 to the current MPEG-4, H.264, there has been a long research history, and many international standards have been proposed, but the basic idea is still block compression and motion estimate. Motion vector prediction techniques are obtained by performing scalar median filtering on motion vectors of adjacent macroblocks. However, in actual video coding, the motions of adjacent macroblocks are not only consistent in the X direction, but also consistent in the Y direction, and they should be inseparable. The present invention is based on such a starting point for motion vector prediction of.

In video coding, the determination of the initial position of the block matching motion estimation, that is, the prediction of the motion vector, determines the performance of the entire coding system. If the predicted motion vector is close to the real motion vector field, then the accuracy of motion estimation is very high, and the residual error between the current block and the predicted block is very small, thus improving the coding efficiency of inter-frame prediction; on the contrary If there is a big difference between the predicted motion vector and the actual motion vector field, it may fall into a local minimum during the motion estimation search process, thus affecting the performance of the entire coding system.

Contents of the invention

The purpose of the present invention is to provide a method for motion vector prediction.

A motion vector prediction method based on vector median filtering, which can predict the prediction vector of the current macroblock or block according to the motion vectors of the surrounding macroblocks or blocks. It first calculates the distance between the motion vectors of the left macroblock or block, the upper macroblock or block, and the upper right macroblock or block, then sorts their distances, and selects the motion vector corresponding to the middle distance as the current macroblock or block The motion vector of the predicted motion vector is centered on the motion search, and the final predicted vector is calculated.

The feature of this method is to first calculate the distance between the motion vectors of adjacent blocks or macroblocks, and select the motion vector of a certain adjacent macroblock or block as the motion vector of the current block or macroblock according to the distance relationship between them . The method mainly includes calculating the distance between motion vectors of adjacent macroblocks; selecting the motion vector of one of the macroblocks or blocks as the predicted motion vector of the current macroblock or block according to the distance relationship between the motion vectors. This method is applicable to any video coding with inter-frame prediction.

A method for motion vector prediction, using the distance between the motion vectors of adjacent macroblocks or blocks as a criterion to select the motion vector of the current macroblock or block, the steps are as follows:

(1) Calculate the distance between the motion vectors of adjacent macroblocks or blocks, respectively;

(2) Sorting and comparing the distances between them, and selecting the motion vector of the block or macroblock corresponding to the intermediate distance as the motion vector predictor of the current block or macroblock.

Computes the distance between motion vectors of adjacent macroblocks or blocks.

Sort the distance between motion vectors and perform vector median filtering.

The motion vector obtained by vector median filtering is used as the predicted motion vector of the current block or macroblock.

With the predicted motion vector as the starting point, motion estimation is performed on the current macroblock or block in the reference image.

Description of drawings:

FIG. 1 is a schematic diagram of motion estimation in the present invention.

Fig. 2 is a positional diagram of the current block and surrounding blocks in the present invention.

Fig. 3 is a schematic diagram of motion vector prediction in the present invention.

Fig. 4 is a schematic diagram of motion search in the present invention.

Fig. 5 is a flow chart of the method for motion vector prediction in the present invention.

Detailed ways

Figure 1 is a schematic diagram of motion estimation. When performing motion estimation on the current sample block with a size of NxN, when the search range is R, the sample coordinate of the upper left corner of the current sample block is taken as the origin, the positive direction of the X coordinate is horizontally to the left, and the positive direction of the Y coordinate is vertically downward, from Select an NxN reference sample best matching block in the range of (-R, -R) to (N+R, N+R), so that the distortion between the current sample block and the reference sample block is the smallest, or when performing rate distortion In the case of optimization, the reference sample block with the highest coding efficiency is selected. The two blocks with thick black lines in FIG. 1 are reference sample blocks corresponding to motion vectors (R, R) and (-R, -R) respectively.

Figure 2 shows the positional relationship between the blocks and surrounding blocks. In the process of motion estimation, the starting position of motion estimation is obtained by predicting the motion vectors of adjacent blocks. In Figure 2, the motion vector of the current block is obtained by predicting the motion vectors of adjacent blocks A, B and C of. If the motion vector of block C is not available, the motion vector of block D is used instead.

Fig. 3 is a schematic diagram of motion vector prediction, in which the X-axis represents the motion vector in the horizontal direction, and the Y-axis represents the motion vector in the vertical direction. Va, Vb, and Vc represent motion vectors of block A, block B, and block C, respectively, and they include the X direction and the Y direction. The length of the line between them represents the distance between the corresponding motion vectors, and then the distance between them is sorted, and the motion vector corresponding to the side of the middle distance is selected as the final motion vector, as shown in Figure 3, for VaVb, VbVc and VcVa are sorted, the side of the middle distance is VcVa, and the motion vector Vb of the B block corresponding to it is the motion vector predictor value of the current block.

The motion vector obtained through the above steps is used as the predicted motion vector of the current block. The motion search process is shown in Figure 4, with the predicted motion vector as the center, and the blocks at each position are sequentially matched according to the sequence shown in the figure. , within a certain range to search for a block that best matches the current block as the prediction block of the current block.

Fig. 5 is the method for motion vector prediction, and its specific steps are as follows:

Step S1, assuming that the motion vector of the left adjacent block A of the current block or macroblock is (Vax, Vay), the motion vector of the upper adjacent block B is (Vbx, Vby), and the motion vector of the upper right adjacent block C is (Vcx, Vcy); their positional relationship is shown in Figure 2.

Step S2, calculate the distance between motion vectors between adjacent blocks according to the following method:

DISTANCE(A, B)=|Vax-Vbx|+|Vay-Vby|

DISTANCE(B, C)=|Vbx-Vcx|+|Vby-Vcy|

DISTANCE(C,A)=|Vcx-Vax|+|Vcy-Vay|

Step S3, sorting the distance DISTANCE (A, B), DISTANCE (B, C) and DISTANCE (C, A) between the motion vectors, and selecting the motion vector corresponding to the intermediate distance as the predicted motion vector of the current block; for example : if DISTANCE (A, B) < DISTANCE (B, C) < DISTANCE (C, A), then select the motion vector of block A corresponding to DISTANCE (B, C) as the motion vector of the current block;

Step S4, taking the predicted motion vector of the current block as the starting point, performing motion estimation on the current block.

Claims (6)

1. A method for motion vector prediction, characterized in that, using the distance between the motion vectors of adjacent macroblocks or blocks to select the motion vector predictor of current macroblocks or blocks as a criterion, the steps are as follows: (1) Calculate the distance between the motion vectors of adjacent macroblocks or blocks, respectively; (2) Sorting and comparing the distances between them, and selecting the motion vector of the block or macroblock corresponding to the intermediate distance as the motion vector predictor of the current block or macroblock. 2. The method for motion vector prediction according to claim 1, characterized in that the distances between the motion vectors are sorted, and vector median filtering is performed. 3. The method for motion vector prediction according to claim 2, characterized in that the motion vector obtained by vector median filtering is used as the predicted motion vector of the current block or macroblock. 4. The method for motion vector prediction according to claim 1, characterized in that, the motion estimation is performed on the current macroblock or block in the reference image with the predicted motion vector as the starting point. 5. the method for motion vector prediction according to claim 1, its concrete steps are as follows: Step S1, assuming that the motion vector of the left adjacent block A of the current block or macroblock is (Vax, Vay), the motion vector of the upper adjacent block B is (Vbx, Vby), and the motion vector of the upper right adjacent block C is (Vcx, Vcy); Step S2, calculate the distance between motion vectors between adjacent blocks according to the following method: DISTANCE(A, B)=|Vax-Vbx|+|Vay-Vby| DISTANCE(B, C)=|Vbx-Vcx|+|Vby-Vcy| DISTANCE(C,A)=|Vcx-Vax|+|Vcy-Vay| Step S3, sorting the distances DISTANCE (A, B), DISTANCE (B, C) and DISTANCE (C, A) between the motion vectors, and selecting the motion vector corresponding to the intermediate distance as the predicted motion vector of the current block; Step S4, taking the predicted motion vector of the current block as the starting point, performing motion estimation on the current block. 6. the method for motion vector prediction according to claim 5 is characterized in that, step S3, is carried out to the distance DISTANCE (A, B) between motion vector, DISTANCE (B, C) and DISTANCE (C, A) Sorting, select the motion vector corresponding to the middle distance as the prediction motion vector of the current block, when DISTANCE (A, B) < DISTANCE (B, C) < DISTANCE (C, A), select the corresponding DISTANCE (B, C) The motion vector of block A is used as the predicted motion vector of the current block.

Images