New Distributed Video Coding Paradigm with Partial Boundary Matching Algorithm, Spatio-temporal Texture Synthesis and Image Inpainting

This paper presents a new paradigm for the Distributed Video Coding (DVC), its coder architecture is quite different from the current DVC video coding architecture. In DVC codec we shift the complexity at encoder to the decoder side, which is suitable for many specific applications such as the low power surveillance, wireless mobile phone, and visual sensor network. Unlike the most popular current DVC coding scheme, named as the Wyner-Ziv (WZ) DVC codec; it divides video sequence into WZ frame and key frame (H.264/AVC Intra), instead we remove the WZ frame and design a new DVC codec architecture associated with key frame (i.e., H.264/AVC intra). Since in our proposed codec it is no need to design any channel coding (e.g. turbo coding or syndrome coding) as in the WZ frame, hence requires less computational complexity and reduces hardware cost of the encoder, simultaneously. Since the WZ frame was removed to avoid the rate-distortion (RD) performance degradation, the quality of decoded video sequences for relative high transmission rate (e.g., great than 350 Kbps) could be improved, while remains having similar RD performance as the WZ DVC codec in lower transmission rate. To verify the merits of the proposed DVC video coding scheme, four popular test video sequences are adopted, under specific imposed constraints. From computer simulation results we show that, in general, the RD performance improvement gain compared with the H.264/AVC Intra frame coding, is in the range of 0.5∼2.3dB.