Research On Scalable Video Coding And High-Performance Video Coding

In the past two decades, there have been fascinating developments in multimedia communication, which is mainly based on video applications. The signigicant advances in digital video compression algorithms increase the efficiency of video delivery and storage continuously. At the same time, the explosive growth in wireless and networking technology has profoundly impulsed the video applications. A series of video coding international standards are established by the ITU-T and ISO/MPEG, in order to meet the increasing demond of various video applications. The latest one, H.264/AVC, is widely accepted and used due to its high performance in both compression and network-friendliness.Recently, the development of video coding reveals two trends. On the one hand, more and more video applications are utilized in a scenario with different network capacities and diverse teminals; therefore, the bitstream should have good adaptivity and robustness. Scalable video coding is an effective solution to the heterogeneous problem. A scalable video coding extension has been add into H.264/AVC, that is H.264/SVC. On the other hand, it can be predicted the future video application will be with higher resolution and quality, thus, the improvement of the compression performance is also needed. The establish of the next generation video coding standard has been carried out by ITU-T and ISO/MPEG.This dissertation focuses on some technologies of scalable video coding and high-performance video coding. The main achievements are as follows:More coding modes are used for H.264/SVC spatial enhancement layer macroblocks; as a result, the encoding computational complexity is a challenge for realtime application. The fast mode decison algorithm utilizes the relationship of the coding modes, quantification parameters between different layers to reduce the number of the candidate modes. Then, an early termination based on the cost relation of neighbouring blocks is used to skip the mode decision procedure adaptively. The experimental results show the proposed algorithm can effectively reduce the encoding time with negligible bit-rate increases.A group of separatable 1D poly-phase 4-tap FIR filters is used in H.264/SVC inter-layer intra prediction. Studies and analysis reveal that this simple scheme is not optimal to process video\image, which is a typical form of 2D non-stationary signal and with unavoidable aliasing artifacts. Two improvd inter-layer intra prediction algorithms using adaptive Wiener filter are proposed. The first one is inter-layer intra prediction with adaptive upsampling filter; the second one is adaptive optimizing filter for inter-layer intra prediction. These algorithms can generate more accurate prediction than the original H.264/SVC inter-layer intra prediction method. Experiments demonstrate that the proposed algorithms can effectively improve the coding performance of the H.264/SVC intra frames.B macroblocks will be used widely in the future due to its sufficient capacity of removing the inter-frame relationship. The proposed algorithm improves the motion prediction in bi-directional coding of B macroblocks. It utilize an adaptive scheme to decide the best motion vector prediction from a set of predictiors, including temporal, spatial and temporal-spatial. It makes the motion vector prediction more accurate; as a result, the residual can be reduced. The experimental results show that the proposed method can increase the compression performance of B frames,especially in the low bitrate scenario.