Fast Algorithm Research On Intra Prediction And Motion Estimation Of H.264Coding Standard

H.264has better performance than previous standards, and has been welcomed in the world. While H.264has excellent compression performance and a good network of affinity, it also brings the shortcomings of high complex computation and larger time-consuming. Therefore, in order to applied in more widely practice, the algorithms of H.264video coding need to be optimized to achieve the purpose of real-time encoding. This paper studies the fast algorithm, focusing on the intra prediction and motion estimation which are the most complex and time-consuming to compute in H.264/AVC core technologies.This paper analyse the intra prediction and motion estimation techniques of H.264coidng standard. Aiming at overcome the high complex computation, large Computing redundancy and long time encoding of intra prediction, this paper proposed two fast intra mode decision algorithms. Firstly, four prediction modes of chroma prediction modes are reduced into one or two modes according to the chroma textural features of a MB. Secondly, the mode with the lowest Sum of Absolute Transformed Difference (SATD) will be selected as the best prediction mode in luma16×16prediction. Then, the number of nine candidate modes can be reduced into three in luma4×4prediction according to the orientation SATD. Experimental results show that the coding time of the proposd method is decreased about68%while maintaining similar PSNR and bit rate. And in order to dividing the sub-block type of intra luma prediction, this paper gives an adaptive mode decision algorithm for intra prediction. Firstly, the algorithm selects the predicion block size according to the texture features of images and a threshold which is changing adaptively with QP. Then, in order to reducing the candidate modes, the most probable mode and two modes with the lowest SATD are selected to replace nine modes of the original algorithm as the candidate modes according to principles of statistics in intra4×4prediction. Experimental results show that comparing with the test mode of JM86, the encoding time of proposed algorithm is decreased about57%while maintaining similar PSNR and bit rate.In UMHexagonS algorithm optimization, this paper combined with the Fibonacci sequence and center-biased feature to improve the search step and search template of UMHexagonS. Firstly, the search step is determined by the progressive relationships of the Fibonacci sequence. Secondly, some search points which lead to computing redundant are deleted. At last, the search template of big hexagon is modified by the center-biased feature. Experimental results show that the new algorithm maintains the same bit and PSNR of UMHexagonS, and reduce the time of motion estimation. And with the improvement of image elements, image complexity and search range, the time for motion estimation becomes less and less. It can be reduced to an average of23.82%of the motion estimation time in the search range of64.