Visual Media Coding and Transmission

4.6 Symmetric Distributed Coding of Stereo Video Sequences

In this section, a new coding scheme based on distributed source coding principles for stereo sequences is proposed. The core of the architecture is based on the pixel domain Wyner–Ziv codec described before and uses rate-compatible punctured turbo (RCPT) codes to correct the side information into the decoded frame. Similarly to the aforementioned works, the proposed algorithm exploits the inter-view redundancy at the decoder only. Despite the previous works, it is fully symmetric, since neither of the two views needs to be chosen as the reference. Also, the quality of the generated side information for different GOP sizes and target distortions is investigated. This work exploits some of the ideas originally developed within VISNETI [26]. Nevertheless, the exploitation of inter-view dependency is novel, as is the study of the behavior of the fusion algorithm as a function of the GOP size and target distortion.

4.6.1 Proposed Technical Solution

Let us consider a scenario with two video sequences, X₁ and X₂, representing the same scene taken from different viewpoints. The coding architecture described in Section 4.2 could be applied independently to each of the two views, achieving very low encoding complexity. Nevertheless, in this case the inter-view redundancy would not be exploited.

Figure 4.28 Block diagram of the proposed ...