The Research Of Key Techniques Of Scalable Video Coding Based On Extension Of H.264

With the rapid development of internet, communication, home network, consume electronics and enlargement of the markets, the multimedia service, especially video stream service has come to be one of the most promising value-added services.Facing the raising complexity of video application environment (heterogeneous networks, fluctuation of channel width, diversity of terminal devices, flexibility of service policy and individuation of request for video from terminal clients) the video coding system has to provide the scalability of temporal, spatial and quality (SNR) with lower complexity. To reply the challenges, the Joint Video Team (JVT) from ITU-T and MPEG begins to design a new video coding framework, which is called Scalable Video Coding (SVC) about at the end of 2005. The emerging SVC prefers the Extension of H.264 to Wavelet as the implement framework. The Extension of H.264, which is proposed by HHI (an institution from Germany), incorporates efficiently the function of scalability of temporal, spatial and quality as well as contains the good encoding performance of H.264. SVC standard will be the first international standard that supports the full scalability. SVC standardization will be the most important project for JVT during the period of 2005~2006. So studying the SVC standard further will make great sense.Based on SVC standard and the Extension of H.264, the paper reviews the history of scalable video coding and deficiency of the existing video standards in scalable video coding, discuss the emerging SVC standard and its key techniques, Extensively studies the pivotal coding techniques of H.264, elaborates the extension scheme of H.264 on temporal, spatial and quality extension and its correlative addition/modification in syntax/ semantic and finally discuss the platform of test software and the experiments on it.To implement the SVC base on Extension of H.264 two class of problems need to be soluted: one is the coding efficiency of H264 core, the other is the efficiency of additional scalability. For the former three new/improved arithmetic are proposed dealing with motion estimation, block mode selection and rate control, for the later one with spatial sampling. 1) Brings forward a new algorithm, which can self-adaptively locate the search-start-dot and search-path. The algorithm is about 10% more quickly than the full searching with almost the same precision.2) Proposes an INTER block mode decision method based on gradient and rate-distortion optimize for H.264, which eliminates some rarely possible modes according to the gradient distribution and reduces evidently the number of modes for the RDO procedure and its time.3) Proposes an improved scheme on rate control on Basic-Unit layer, which doesn't only take account of the relative MAD-complexity of Basic-Unit (BU) but also considered the scene-change. The tests prove that the improved scheme on Basic-Unit layer can predict MAD more accurately and get better subjective quality.4) Proposes a new sampling method based on DCT which can improve the performance of spatial scalability and the PSNR by 1~3 dB compared with standard method of SVC.