Research On Inter-Layer Prediction For H.264 Scalable Video Coding

With the advances in video coding technology, as well as the development of network infrastructure, storage capacity and computer performance, there is a significant increase in video applications. Meanwhile, as the enhancement of the progress of spiritual civilization construction, communication with the usage of images and video as the carrier of information has become increasingly universal. Usually, the same source content is accessed by a large number of terminals with different performance, from mobile phones, with small screen and limited processing power, to personal computers, with high-definition display and excellent performance. Therefore, the server needs to provide video bitstream coded in different image sizes, different frame rates and different bit rate. Scalable video coding (SVC), an extension of H.264/AVC standard, can provide a scalable video stream, achieved by temporal, spatial and quality scalable coding mechanism, which can meet the needs of different terminals existing in the heterogeneous network.Firstly, this paper introduces the scalable video coding scheme that some bit of information of the coded bitstream can be removed to meet a variety of requests from end-users with different capabilities or to adapt to the heterogeneous network environments. Chapter 2 give some specification about the principles of H.264/AVC codec, as well as its key techniques including prediction coding, transform coding, entropy coding and deblocking filter.This paper goes on with details of scalable video coding (SVC) conducted through temporal, spatial and quality scalability: the hierarchical B structure used in temporal direction can improve coding performance; inter-layer motion and residual prediction can improve the performance of spatial and quality scalability; the usage of key frames can effective control the drift effect in quality scalable coding; single motion compensation loop decoding strategy reduces the complexity of decoder.Finally, simulations are conducted to analyze the performance of SVC. Through the analysis of coding performance of the traditional inter-layer intra prediction, which is implemented by a simple interpolation filter, an improved algorithm for inter-layer intra prediction is proposed, which is based on the quadtree structure. This algorithm using an adaptive Wiener filter to rcover the reconstruction image of base layer achieved by upsampling operaltion. Moreover, the filter is redesigned only to region where the ?ltering operation is actually needed. The agorithm improves the prediction accuracy between spatial layers. Finally, simulation results of this adaptive filtering algorithm are analyzed.To further improve the performance of intra prediction and compensate the defect of independent usage of spatial or temporal prediction, an method based on the concept of motion search algorithm is proposed to improve the accuracy of intra prediction. Finally, experiments are conducted to analyze its performance.