Research On Distributed Video Coding

With the development of Wireless Multimedia Sensor Networks (WMSN) and the emergence of smart phones, wireless and mobile become the trends of communication development. Distributed Video Coding is a coding scheme which is suitable for video communication in wireless networks. It has a feature of low-complexity encoder and high-complexity decoder. To reduce the complexity of encoder, it transfers the motion estimation and motion compensation which belong to encoder in traditional video coding to decoder. Because the side information directly determines the quality of recovery frame, and the rate control determines the size of transmission bit rate, the side information and rate control is the core of Distributed Video Coding.Firstly, this paper in-depth studies two algorithms about how to generate the side information (SI):motion compensated extrapolation (MCE) algorithm and motion compensated interpolation (MCI) algorithm. MCI can reflect the change of current motion vectors, but it failed to use time correlation of previously decoded frames to maintain the continuity of the motion vectors. However, MCE maintains the continuity of motion vectors, but it failed to reflect the suddenly change of current motion vector. This paper combine MCI with MCE, this new algorithm can not only maintain the continuity of the motion vectors, but also reflects the change of current motion vectors.Secondly, in the proposed MCI-MCE algorithm, this paper uses the overlapped block motion compensation to generate SI. This paper proposes to use the overlapped block motion compensation to handle the overlapped projection which is from the front frame to the current frame. On the one hand, the proposed algorithm solves the block effect which is caused by the difference of motion vector within a block when the block at the boundary line of the background and object. On the other hand, to the original side information generation algorithm using the average method to handle the overlapped projection, the proposed algorithm improves the Smoothness of the blocks.Finally, this paper proposes the combination of frame levels and block levels to reduce the rate of channel. Combining zero motion blocks B-Skip algorithm and frame F-SKIP algorithm, this paper use the adaptive Laplace factor to control rate when it estimate the factor of virtual channel model (VCM), which depend on the size of motion vector.