Research On Motion Estimation And Intra-Inter Prediction Algorithm For H.264 Standard

With the era of information and the rapid development of computer and multimedia technology, application area of digital video unceasingly extends. However, video data quantity is very huge and brings enormous trouble to image tranfer and storage. So, video compress algorithm that is more efficient can solve the problem of video data transfer and storage. H.264 is currently the newest video compress standard. Compared with other video compress standard, H.264 has higher video compress performance. For example, under the same condition of coding quality, H.264 bitrate required is about 51% of H.263 and 61 % of MPEG-4. But improving compress rate is at the cost of increasing coding complexity. Research shows the calculation time of H.264 inter and intra coding makes up 85% of the whole coding. Thus, improving intra and inter coding speed has an immense significance to real-time coding video image.The paper does research for H.264 inter and intra algorithm and proposes a new sub-pixel search algorithm, an improved UMHexagonS algorithm, sub-pixel full search algorithm and fast intra mode seletion algorithm. First, the new sub-pixel search algorithm has been proposed on the basis of relevance between integer-pixel and sub-pixel. Experiment shows, compared with sub-pixel full search algorithm, the algorithm reduces search points by 50% and the image quality reconstructed is much closer to full search algorithm. Second, because search scope of UMHexagonS algorithm is invariable, if we can dynamically adjust search scope and keep image quality, we can reduce research points and improve algorithm speed. Improved UMHexagonS algorithm is proposed on the basis of this thought. Study shows that under the situation of little influence on image quality, proposed UMHexagonS algorithm decreases search points, reduces compute quantity and improves efficiency of motion estimation. Third, due to this reason that the points from H.264 sub-pixel full search algorithm are the local best ones, the paper proposes an improved algorithm. Improved algorithm searches all possible points, prevent to get local best points and effectively avoid the disadvantage of H.264 subpixel full search algorithm. Study proves that new algorithm can improve the image quality more than H.264 algorithm.