Research On Fast Motion Estimation Technology Based On H.264 Standard

VCEG(Video Coding Experts Group) and MPEG(Moving Picture Experts Group) published H.264, which is a new standard for video compression coding. This standard can obtain excellent coding efficiency and performance in the low bit rate environment. So it can meet the requirement of the network video with low bandwidth and high quality. But, compared with most of the old eneoding stnadards, the complexity degree of H.264 eneoding have been increased greatly through the application of advanced algorithms. It becomes a crucial point in H.264 research that how to reduce the complexity of encoding and the better application of H.264 in practice.Motion estimation is the key technology of video compression, which dominates the performance of compression and the time for computation. So it is mainly studied to search a brief fast efficient algorithm of motion estimation for video coding. In H.264, the variable-size block motion estimation is adopted for inter prediction, and in this paper, parts of its algorithms have been studied and improved. The specific contents are as follows.Firstly, high performance block matching motion estimation algorithm is studied. The basic principles of typical motion estimation algorithms are analyzed, and their advantages or disadvantages have been tested by experiments.Secondly, an efficient searching algorithm named Adaptive Diversity Search Strategies (ADSS) is proposed. ADSS can produce better motion estimation results but does not increase the searching computation. It has a wider scope of application.Thirdly, the block merging and splitting algorithms are studied. An advanced algorithm based on merging and spliting procedures is proposed. The algorithm is helpful to find the motion vectors of merged or splitted blocks in a more efficient and accurate way.Fourthly, the inter-mode decision algorithm has been optimized. The algorithm is improved by the application of block merging and splitting judgement. The improved algorithm can reduce the computation of block mode decision considerably.Finally, experiments are made to prove the performance of these new algorithms. The results of experiment show the performance of new algorithms are much better than other algorithms'. The expected goals of study are realized.