The Research Of Motion Estimation Algorithm Based On H.264/AVC

H.264/AVC (Advanced Video Coding) only requires that the codec syntax and framework, not to restrict the concrete realization of the coding module. This will not only ensure the compatibility, but also to encourage the research and creation on the basis of its standard. Fast motion estimation algorithm UMHexagonS (Unsymmetrieal-CrossMuti-Hexagon Search) is an excellent algorithm based on H.264/AVC. However, the sensitivity and adaptability of the algorithm in the vertical direction is still not perfect. Besides. UMHexagonS as a hybrid fast motion algorithm, its encoding efficiency can be further improved.1. A systematic analysis of the key technologies for video encoding and H.264/A VC video coding standard, and in-depth study on UMHexagonS algorithm. Founding the truth that there is a space for improvement in speed of motion estimation and coding efficiency of UMHexagonS.2. Researching a variety of estimation algorithm and reading a lot of papers. Adding8points into Stepl (asymmetric cross-type search) of UMHexagonS to improve the adaptability of its movement in the vertical direction. Setting up a parallel algorithm for Step3(Uneven multi-hexagon-grid Search) of UMHexagonS to further improve the performance of the algorithm for motion estimation.Besides. in order to improve the coding efficiency of the algorithm and reduce motion estimation time. Making the first small diamond-shaped contour search and then large diamond-shaped contour search for Step2(small rectangular window search) of UMHexagonS. Putting forward three search method to replace the original Step4-2(small diamond search), and the original Step4-2is just a Jump object for the most satisfied Early-termination.3. The improved algorithm in C language is integrated into the H.264/AVC reference model JM10.1and then compiled and tested with VC6.0. The results show that the improved algorithm has more lower bit rate and more lesser motion estimation time comparing with UMHexagonS. while its reconstructed image quality and PSNR (Peak Signal to Noise Ratio) are equivalent to oneFinally, this paper proposes the next step could be to study, they are starting search point prediction optimization, intra-prediction mode selection. bit rate controlling and the embedded transplantation and optimization of algorithm.