Study On Scalable Video Coding And The Coding Techniques For Its Reliable Transmission

Due to the explosive growth and great success of Internet, the requirements of network multimedia services have been increased extremely, e.g. network VOD, IPTV, etc. and been widely considered by industry and academia as the next growth point of service and economy of information industry. However, the Internet, which was is based on IPv4 and initially designed for the data transmission among computers for research institutes and universities, is inherently a heterogeneous and dynamic network, cannot guarantee necessary QoS. Therefore, it is necessary to provide relative technical support at transmission and application layers. However, the complicated transmission characteristics of the IP network present great challenges for efficient and reliable video stream transmission. So, the researches on the video coding techniques, which are suitable for transmission over IP network, have theoretic and application significance, and become one of the most attractive directions in video coding fields.The work of this dissertation is concentrated on the video coding techniques, which support reliable video transmission over IP network, so as to enhance the coding efficiency and error resilience, and improve compressed image qualities as well. Firstly, the fine granularity scalable (FGS) coding is chosen as the primary research goal. Secondly, the interleaving-based video multiple description coding (MDC) of hybrid coder is studied. Lastly, post-processing for compressed video, especially, the deblocking filtering for hybrid coding, is studied.In the study of FGS, an abstractive data structure description is presented at first to represent coded streams. Accordingly, several current scalable coding schemes are categorized into three types, and three well-known FGS algorithms are described and analyzed briefly. Then, the MPEG-4 FGS is tested with standard test sequences in details. Based on the analysis of the experimental results, two aspects to improve FGS coding efficiency are proposed. One of them is to use stream switching to adapt to wide range bit rate variations, and the other is to improve the enhancement layer coding directly. At last, three possible ways to improve FGS enhancement layer coding are discussed, such as using morphological and sub-band data re-arrangement of the enhancement layer, as well as direct coding of the least significant bit plane. Experimental results show, that by DCT coefficient re-arrangement, the coding efficiency of FGS is improved effectively, and the MPEG-4 advanced features FW and SE are realized inherently without any extra bit rates.In the study of multiple video stream switching, the principles of stream switching are described at first. Based on the analysis of the encoder configurations of current well known stream switching schemes, the core of multiple stream switching is discovered, that is to insert suitable"anchor points", the switching frames with proper quality in the target bit streams. According to the different requirements for the stream switching up and switching down, a novel full embedded stream switching scheme (FESSS) is proposed. Details of the design procedure of the new scheme are described, and two corresponding realizations of the algorithm for generating basic switching stream are presented. The merits of the new algorithm are to generate basic streams by open-loop manner, therefore, multiple streams can be generated in sequence by only one encoder, and the computational complexities are deceased obviously. Experimental results demonstrate the coding performance and the practical significance of the new scheme.In the study of interleaving based multiple description coding, strategies to improve reliability of network video transmission with error coding control are described at first. The error resilience coding, as the one important aspect of error coding control, are discussed briefly. The differences of two important error resilient coding, the scalable coding and the multiple description coding, are analyzed. After explaining current known MDC of hybrid coder from a new point of view– data interleaving based technique, the study is concentrated on the simple and effective MDC that use pixel interleaving. Based on the brief description and analysis of ORB-DCT algorithm, a new scheme is proposed. Details of the new algorithm are presented, which possesses clear physical meaning and is suitable to use fast DCT to improve processing speed. Then further studies are carried out from a general anti-aliasing point of view. As result, the sub-sampling patterns are expended from 1D to 2D, and the asymmetric feature in images is found from experiments. At last, explanation on the interleaving-based optimal reconstruction is made from a new point of view, which provides a possible direction to expend MDC from two descriptions to 4 descriptions, and to find their optimal reconstruction as well.In the study of post-processing for compressed video, the deblocking filtering based on soft-thresholding on wavelet domain is discussed in details. Based on the analysis on the problems of current algorithms, possible direction to improve the estimation of threshold values by using non-texture areas in compressed image are proposed, and two corresponding algorithms DCCST and VFST are presented, which improve both of the threshold value estimation and thresholding filtering processing. Special considerations are also taken for compressed video as well. Experiments are carried out both for the compressed images and videos. Comparing with the known algorithm, the test results demonstrate that the new algorithms can make more precise estimation of the threshold values, and improve deblocking obviously, while the computational complexities are decreased effectively. At last, from the analysis of the test results, a new hypothesis of the consistency of the threshold values is proposed and verified with experiments. This hypothesis is significant for improving the processing speed in real time deblocking filtering.At last, the summary of the dissertation is made, and the directions for further studies are presented.