Channel division based on foreground segmentation in multi-view distributed video coding

In this paper, a channel division based on accurate side information generation method is investigated if it can improve the coding efficiency of multi-view distributed video coding (MDVC). Each block in a side information frame is taken among those of inter-view or intra-view key frames based on foreground segmentation, and the characteristics of virtual channel noise for the generated fused side information is investigated. Based on the magnitude of average SAD of each block, a block is assigned to a certain channel, and Laplacian channel parameters for each channel is estimated for MDVC decoding. Simulation result shows that side information fusion based on foreground segmentation helps to improve coding efficiency in MDVC. However, channel division does not always enhance the coding efficiency, but it should be designed considering the level of channel errors and the accuracy of side information.