Research On Optimization Techniques In Intra Prediction Of Video Coding

The video coding technology has never stopped advancing to satisfy various demands on video appliances under different circumstances. Requirements for higher video resolution, high definition and super high definition video appliances as well as the widespread use of mobile videos all claim more on video coding standards which are supposed to achieve better coding efficiency and stronger robustness. Study on optimizing techniques of video coding plays an important role in answering the call for such higher requirements. Intra prediction, as an important part of coding technology, employs many complicated coding techniques. Especially, the rate distortion which needs much computing adds to the overall computing complexity. As a result, it has been a heated research field to decrease the complexity of intra prediction while guaranteeing the implementation quality of video coding standards. This work is to explore the optimizing methods on intra prediction for the popular H.264/AVC coding standard and the recently raised high efficiency video coding standard (HEVC).As for the problem of intra prediction in H.264/AVC that the DC prediction mode fails to exploit the special relativity, this research puts forward the iterative prediction method. Through the study of directional features and texture characteristics in local areas of the prediction, horizontal and vertical prediction strategies are introduced. The iterative predicting method only uses neighboring reference pixels from left and up coded pixels, and the reference pixels all have the same prediction weight. The results of experiments verify that the iterative prediction method promotes the coding efficiency of DC prediction mode and it can be easily implemented.As one of the most frequently studied intra prediction optimizing techniques of H.264/AVC, the fast mode selection algorithm is further improved. The technique features of the existing fast algorithm of mode selection are examined and a simplified comparing method is generated concerning these. That is, reduce the coding time while reserving the performance of the fast algorithm. So the hybrid fast mode decision is introduced and the filter designed corresponding to each predicted direction and mode. This filter efficiently ends the depending relationship between directional prediction modes and also absorbs features of H.264/AVC intra prediction so that a better performance can be achieved than the existing ones.In this research, fast mode selection based on local information is proposed. The capability of mixed fast mode decision algorithm is tested and relative refinements are made. The data processing approach of the algorithm is improved so that the computing complexity is further lowered. Information of the local area is generated and the more appropriate generating method is designed according to data characteristics of the filter group. Therefore, the coding performance is sublimed with less coding time. Based on the mode decision hierarchy, synthesized optimizing method is introduced by integrating with block size type decision method which can further decrease the coding time.With the analysis of characteristics of HEVC intra prediction, the dissertation puts forward the adaptive fast mode decision algorithm. It has most probable modes as references and adaptively chooses from the prediction mode candidates based on calculating results of most probable modes at the rough prediction mode decision stage. This method largely cuts down the computing complexity with little impact on the coding performance.