Study And Optimization On Scalable Video Coding

Network oriented video service applications have a rapid growth in recent years. Video has been increasingly adopted by a wide range of applications. The multiformity of the user facility, the wide diversity of video applied environment and the time-varying network bandwidth, all of these make it hard for the traditional single speed video stream to meet the needs of terminal diversity, heterogeneous network and the varying bandwidth. So video service system requires an encoding system which can adjust different kinds of network environments and user terminals. Scalable video coding has a property of encoding once while decoding many times, which provides a new idea to solve the adaptability of video streams, so SVC becomes a research hotspot of video coding.In this paper, I conducted an in-depth study on both H.264-based and wavelet-based scalable video encoding algorithms, analyzed the encoding principle and implementation skeleton of the two systems, compared the difference in performance between them. On this basis, the thesis optimized the H.264 SVC encoding scheme which had obvious advantages, mainly including the following two components:(1) The encoding optimization of H.264 SVC. The original encoding algorithm used rate-distortion optimization to determine macro-block coding mode, and this full traversal strategy made motion estimation for different partitions to pick up the optimal mode, which tremendously increased the computational complexity of encoder. The biggest difference between SVC and AVC is the characteristic of multilevel coding. Each space layer image has inter-layer correlation. In this paper, taking advantage of inter-layer prediction information, I proposed a statistics-based fast mode decision algorithm, which greatly cut down the macro-block mode decision time and improved the coding efficiency.(2) The decoding optimization of H.264 SVC. SVC code stream has space, time and PSNR scalability, so the decoder complexity is comparatively higher than the single layer decoder, which is hard to meet the need of real-time. This paper made an optimization for decoding algorithm of H.264 SVC, with the use of a combination of several modes of operation, such as program code optimization, processor optimization and algorithm improvment, improved the decoding speed efficiently.Through the optimization study, results shows that the encoder coding rate is doubled, and the decoder can achieve the real-time decoding of 4CIF video on PC plantform.