Research On Mode Selection And Motion Estimation Algorithm Based On H.264 Standard

H.264/AVC(called H.264 as follow for short) is a new international standard for video coding, which was published by JVT who was composed of Video Coding Experts Group and Moving Picture Experts Group. Prediction modes selection and motion estimation are key technologies in H.264, they are the main computational cost of video compression and determine the encoding speed and compression quality. Thus, they are mainly studied in video encoding field. To search a efficient video compression algorithm, this paper researched intra-frame multi-mode selection, inter-frame multi-mode selection and multi-reference frames motion estimation algorithms in H.264, and then, improved them.Firstly, the technology of multi-mode prediction coding for intra-frame and inter-frame was studied. According to the direction of the prediction models, a variety of patterns were classified, the scope of models selection was narrowed to improve the algorithms in mode selection technology for intra-frame prediction coding. Then, according to the detail characteristics of the video image, a new mode selection algorithm for inter-frame prediction coding was proposed. Improved algorithm classified the prediction modes by size and determined which type to be used according to the levels and the direction of video sequences to reduce the number of prediction models.Secondly, the study on block-based multi-reference frames motion estimation was made. A multi-reference frame motion estimation algorithm was improved with the use of the space-time relation in video image information. The improved algorithm selected adjacent frames and the best reference frame of the adjacent coded blocks as the reference frames of the current block, so that it could reduce the number of reference frames. At the same time, the algorithm combined a variety of templates to search the reference frames in order to reduce the search points and improve coding dfficiency.Finally, system simulation experiments of the improved algorithms were made. Based on the experiments running, the performances of the algorithms were analyzed and evaluated.