A Multi-view Video Coding Method Based on Distributed Source Coding for Free Viewpoint Switching

In this paper, a multi-view video coding scheme based on distributed source coding theorems is proposed for free viewpoint switching in multi-view video scenario, which can let users freely switch between any two adjacent views even if the switching point is not an intra frame. In this scheme, multi-view video frames are classified into three categories: Intra frames, Predicted frames and Wyner-Ziv frames. Intra frames are encoded and decoded using a conventional intra frame codec. Predicted frames are encoded and decoded using a conventional inter frame codec. Every Wyner-Ziv frame can serve as "switching points", which is encoded using DSC based on channel codes. When Wyner-Ziv frames are encoded, both the temporal correlation and cross-view correlation are considered to generate enough syndrome bits for reconstructing the frames according to different side information obtained at the decoder. The experimental results show that the proposed scheme can achieve better compression efficiency while keeping the free switching property.