Partial Boundary Match Algorithm and Spatio-Temporal Texture Synthesis in Distributed Video Coding

The Stanford's WZ codec, Berkeley's PRISM and Europe's DISCOVER, are considered to be three early major research groups in the distributed video coding (DVC) with better performance. In this paper, we propose a new architecture of DVC, which does not follow the abovementioned solutions. Basically, we adopt the PBMA (partial boundary matching algorithm) and ZMVR (zero motion vector replacement) at decoder to replace the motion estimation at encoder. To achieve the desired performance, in other parts we use the STTS (spatio-temporal texture synthesis) and image inpainting for further processing. From experimental results we show that 4dB (at 300 kbps, PSNR is 34 dB) gain higher than H.264/AVC intra-frame video coding approach could be attained.