| The newest video coding standard H.264has got widely applied because of its excellent performance, which was published by JVT (Joint Video Team) comprised of ITU-T and ISO/IEC. Not only it improves video compression efficiency, but also enhances data transformation in low bit rate network, especially suitable for low broadband and high quality network application. After the analysis and research of H.264encoder, we find motion estimation has become one of the core techniques in the whole video encoder, and the adoption of both the motion estimation and motion compensation technology can eliminate temporal redundancy between video perfectly and improve encoding efficiency consequently. However, the computation in motion estimation is so large that it almost takes over60%-80%of the entire encoding time, which blocks the real-time requirement seriously. Block matching algorithm takes the majority in motion estimation, so in order to make motion estimation algorithm become faster, stronger and more effective, how to do some optimization of these algorithms to get encoding efficiency improved has been a hot spot in video compression area.This paper firstly introduces the basic principle and key techniques of standard H.264and then makes analysis of some typical block matching algorithms including their advantages and disadvantages. It also gets further research of the kernel algorithm named UMHexagonS adopted in H.264, and proposes three aspects optimization according to the existing deficiencies, which contain new search pattern based on distribution of motion vector, adaptive search range method and spiral subset matching strategy. The optimized UMHexagonS not merely can make sure the location and size of the reference search window quickly and efficiently but also gets searching points decreased apparently which still guarantees the good quality of video, and achieves preferable optimization effect.This paper implements the optimized algorithm in the reference model of H.264compiled by VC6.0, and chooses six typical video sequences to test. Experiments show the optimized UMHexagonS gets significant improvement compared to original one. On the precondition of similar reconfiguration video quality and transmission bit rate, the optimized algorithm has motion estimation reduced by13.792%compared with UMHexagonS, and it gets better optimization to video testing sequences with complex movement. Accordingly, the optimized UMHexagonS does can raise motion estimation efficiency and improve the video coding in real-time application. |
