Study On Motion Estimation Algorithm Based On H.264 Standard

Studies on the motion estimation algorithm based on the new video coding standard H.264 is one of the hot tasks in the present research area of H.264. Based on the further study of the characteristics of the H.264 standard and the principle of motion estimation and motion compensation, aimed at the shortcoming of the preferred fast motion estimation algorithm—UMHexagonS algorithm in the reference software models of H.264 and referred to the ideas and technologies used in the existing motion estimation algorithm, this paper first proposes an improved UMHexagonS algorithm and then proposes a novel Predictive Cross-Octagon Search (PCOS) algorithm based on the further analysis and summarization of the excellences of the existing motion estimation technique. The experimental results indicate that these two algorithms can save about 4.7% and 9.9% of the encoding time separately.In sum, this paper is developed according to the following topics:(1) We study the effect of the corresponding parameters of multi prediction modes search, multi reference frames prediction, rate-distortion optimization mode selection, search range and motion estimation algorithm on the performance of the entire encoder. Then, by performing a series of experimental analyses, the author summarizes a parameter configuration scheme, which can make sure that the encoder would have better performance. It has an important guiding effect on engineering practice.(2) We simplify the unsymmetrical-cross search pattern, spiral full search pattern within5×5area around the search center and multi sixteen points Hexagon pattern in UMHexagonS algorithm that have an impact on the search efficiency. The author proposes a novel octagon search pattern that has symmetrical search points located on one circle and use it as our basic search pattern of multi search. Finally, it forms an improved UMHexagonS algorithm—UMOctagonS algorithm. The realization of the algorithm and the experimental analysis proves that the improved algorithm achieve the same PSNR and bit-rate with a certain complexity reduction.(3) By analyzing a lot of motion estimation techniques and by analyzing the characteristic of the real world video sequences statistically, we summarize several key technologies that could improve the motion estimation efficiency. These technologies are the cross search pattern based on the center-biased characteristic of motion vectors, the initial search point prediction based on the temporal correlation and the spatial...