Study On Video Coding Algorithms Of MPEG-4

Recently, due to the fast development of computer techniques, network techniques and imaging processing techniques, real-time video processing technique is widely used in video conference, remote teaching, real-time monitor, etc, which has higher requirements of transmission rate and video quality. Considering limited channel bandwidth and real-time processing requirement, it is necessary to apply efficient video source code with high compression ratio.There is strong relationship between a frame and previous frames in one video sequence on the timeline, so it would achieve high compression ratio by using motion estimation techniques to reduce redundancy. Sprite is an image constructed from video clips and is also a medium for multimedia application, which is one of the core techniques for compressions in MPEG-4 standard using global motion estimation. The main researches in this dissertation are illustrated as follows:(1) After studying the principle of MPEG-4 coding, motion estimation and several common fast search algorithms, Adaptive Cross-Based Search Algorithm (ACBs) is proposed. Making full use of the motion vector's characteristics which are center biased distribution, vertical and horizontal directions, temporal and spatial relevance, the proposed algorithm consists of three search patterns:small-diamond pattern, horizontal-cross pattern and vertical-cross pattern. The algorithm first sets the threshold for stationary block to directly suspend search. Then according to the type of motion adaptively choose the search strategy and the original search point, at the same time selecting the search pattern in accordance with the direction of motion vector. It is proved that the proposed algorithm is superior to the common fast motion estimation in efficiency, such as TSS, DS, and the searching accuracy close to DS algorithm.(2) After investigating the principle of Sprite and global motion estimation, the algorithm is proposed to improve the global motion estimation in MPEG-4 VM in two aspects:foreground object removal and calculated point selection. Each pixel-value of the macro block on sprite is iteratively updated by the value with maximum appearance probability on temporal and spatial distribution. In the result, the generated sprite preserves the complete scene of background and has higher image quality. When selecting the calculated point, first remove the pixels that have no contribution or less contribution to GME according to error function, and then choose the feature set for GME.(3) Finally, it is time to validate that the proposed algorithms are feasible and advantageous. Experiments results show applied with the proposed ACBs motion estimation algorithm and Sprite algorithm, it can successfully realize MPEG-4 coding, with high-quality and high revolutionary. In addition, ACBs gains the obvious improvement in time and search points and the improved Sprite algorithm also can increase computational efficiency.