Research On Spatial Scalable Video Coding Based On H.264

Recently, with the development of coding techniques, applications of video compression and transmission become wider and wider. Different compression and transmission systems have to be constructed for users in different conditions if they are built on the base of traditional non-scalable video encoding techniques. With the manifold of video application, the rebuilt of compression and transmission systems is a huge waste of resource. Therefore scalable video coding techniques are desired for coping with different application.For the need of scalable video coding, some of prior video coding standard has already include somewhat scalability, but its performance turn out to be discontented. H.264 is a classical video coding standard, several coding techniques enable it a good performance. More and more video has been or will be code in H.264 video coding standard. To meet the need of high performance scalable video coding, MPEG decided to draw a new scalable video coding standard based on H.264 on Oct.2004. MPEG and VCEG agreed to made scalable video coding standard an extension of H.264 on Jan.2005.Now, mobile video has become a fashion. Video conferencing and video surveillance have a rudiment, thus all that multimedia will bring up massive data throughput. The good performance of scalable video coding will have a good future. Among of the type of scalable video coding, the spatial scalable video coding is the base of the others. It motivated us to study on the spatial scalable video coding.This thesis first provides an overview of the video coding techniques and the international video coding standard, especially on the key techniques of H.264. Then we analyze the scalable techniques of H.264, especially for the inter-layer prediction in spatial scalability, studies inter and intra inter-layer prediction with detailed performance statistics, and reveals the necessity of optimization of inter-layer prediction. Based on the analysis of the results of experimentation, finally, an optimization method based on the mode of the macroblock is proposed by exploiting the motion, the coefficient values of transform. Experimental results show that with the negligible loss of the encoding performance, the proposed method can reduce the encoding time 64% on an average and shows a good adaptability.