Based On The H. 264 Scalable Coding Algorithm Research

Salable video coding (SVC) is an extension of H.264/AVC. In SVC there are threekinds of scalability which are temporal, spatial, and quality scalability. SVC can provide aflexible solution to cope with the diverse need for heterogeneous transmission systems andvarious terminal equipments. By using inter layer prediction the coding efficiency of SVCis higher than previous standards. However it also leads to high coding complexity andcomputation complexity. So how to decrease the coding complexity and save encodingtime is one of the hot topics.This paper studies the coding principle and some technologiesof SVC with a lot of experiments and promotes several improved solution and algorithms.This paper analyzes the correlation between B pictures of different layers and thecorrelation between B pictures and key pictures in the hierarchical B-Picture structure. Inorder to reduce the large peak signal-noise rate (PSNR) fluctuations, this paper proposes anew quantization parameter allocation strategy which based on the analyzed correlationand combines the persistence of vision.Because intra prediction mode decision is correlative between enhancement layer andbase layer in SVC and the nine intra4×4modes are also correlative, this paper analyzes thecorrelation and proposes a new intra mode decision algorithm which is used inenhancement layer. In this algorithm the intra mode in enhancement layer (EL) is decidedbased on the mode in the base layer (BL). And the number of intra modes in EL is reducedaccording to the texture of frames. Experiments show that the proposed algorithm can savemore than20%encoding time on average with neglected PSNR change and small bit-rateloss.A fast inter mode decision algorithm was proposed in inter prediction procedure. Withexperiments this paper analyzes the inter prediction correlation between spatial layers. Inthe proposed algorithm inter mode in BL is used when deciding the inter mode in EL. Andin order to reduce the complexity of computation the inter-layer residual prediction (ILRP)is dynamically chosen according to the rate-distortion cost (RDCost) of inter mode in BL.Experiments show that the proposed algorithm can save much more encoding time andeffectively decrease the coding complexity.