Researches On Error Control Of Scalable Video Coding And Multi-view Video Coding

In recent years, with the fast development of networking and diversified multimedia application, research on video coding technology is not only involved with the compression characteristic. Heterogeneous network structure and various user terminals require different video resolution, frame rate or quality, these lead to great challenges in the work of video coding standards. Scalable Video Coding(SVC) is developed by JVT(Joint Video Team) of ISO/IEC and ITUT VCEG as an International Standard. At present it is one of the best solutions to the above problem. It can provide scalabilities on temporal, spatial and quality. Multi-view video coding is another video coding standard under development. Multi-view provides better visual experience, and it has high correlation between different views, which makes it possible to compress multi-view video efficiently.. Since degradation, bit error, and packet loss usually happen in channel transmission, and these errors can be propagated to other video frames so as to seriously damage the video quality, error control turns out to be an important technique. Thus study of error control methods in Scalable Video Coding and Multi-view Video Coding is a hot spot in video processing and communication research area.Based on the framework of SVC and MVC standard, this thesis investigates several error control techniques in them. The main contents and novelties of this thesis are as follows:1. Based on redundant picture, this thesis proposes a new error resilience approach to SVC.The redundant picture is an effective method dealing with packet loss in transmission, also it is an important error control technique in video coding. Adding redundancy to the video bitstream may enhances robustness of the transmitted video, but it also will reduce the efficiency of video coding. So, how to appropriately generate redundant pictures and assign them is important. Taking into account of the unsymmetrical protection mechanism in SVC video data transmission, this thesis proposes a new error resilience approach which adds simple exactly copy pictures to the key pictures in enhancement layers. Redundant picture can stop error propagation and improve error resilience without adding too much complexity to the encoder.2 This thesis proposes a new error control approach on MVC based on disparity map. The proposed approach utilizes the motion information similarity between different views to generate disparity map. Then this disparity map is packed into video data stream. At the decoder side, according to the disparity map information, motion information of the lost picture is restored and used to utilize the temporal correlation to generate the lost pictureA number of experiments have been done in this thesis to investigate the proposed error control techniques of SVC and MVC. Experimental results show that, the proposed error control algorithms can improve both objective quality and subjective quality of video compared with traditional methods.