Research On Fast Interframe Coding Algorithms Of Video Coding

Since H.264 was setted as video coding standard in 2003 by the International Organization for Standardization,it is regarded as the next generation of video coding standard for its better performance comparing with former coding standards.H.264 can save at least a half of bit rate in the same image decoding quality.Many advanced technologies are adopted in H.264,such as stratification coding structure which includes Video Coding Layer(VCL)and Network Abstraction Layer(NAL),block-based motion compensation hybrid coding framework,multi-mode and high precision motion estimation,advanced intra prediction,integer transform and quantization.The coding efficiency are greatly improved by these new technologies,but it is at the cost of increasing coding complexity.This paper mainly studies two time-consuming parts in H.264 encoder,the motion estimation and mode selection.Firstly,the basic principle of motion estimation is introduced and in this part we focus on introducing the UMHexagonS fast search algorithm.There are some deficiencies in the UMHexagonS algorithm such as the fixed templates and fixed search steps,and a fast motion estimation scheme using adaptive search templates based on the motion direction and the motion degree is proposed for motion estimation.Secondly,all zero block detection and mode decision theory is introduced in this paper and the threshold of all zero block detection is deduced based on the integer transform and quantization.After analyzing the cost of coding one macroblock,a fast mode selection algorithm based on directional texture is presented for mode decision.Combining the improvements of the two parts which are made above,a fast inter frame coding scheme including both adaptive directional motion search and fast mode selection is proposed at last.Finally,experimental results tested in reference software JM10.1 show that the proposed fast coding scheme can reduce encoding time by an average of 53% comaring with the UMHexagonS while keeping the same bit rate.The proposed fast coding scheme can also reduce encoding time by an average of 47% comaring with the other published algorithms.