Distributed Video Coding with Background Modeling for Wireless Video Surveillance

Video compression takes a very important part in multimedia communication. New applications require constrained resources in terms of power, computational complexity, memory and robustness. Distributed video coding (DVC) provides a solution to achieve such constraints. However, DVC has a low coding efficiency, the overall performance is lower than the H.264/AVC [1] codec. In low-motion scenes such as video conference or video surveillance, we can reduce the storage and transmission cost by using the long term background characteristics. In this paper, we propose a new DVC based on the background modeling of the scenes. We further analyze and design the corresponding optimized structure. We test different low motion video surveillance sequences, and show that the proposed block-based coding paradigm has a better performance compared with current DVC system. Experimental results show that the proposed background modeling method and coding scheme can save more than 14% bitrate and the overall rate-distortion (R-D) performance is increased 1.22 dB in average vs. Low-complexity DVC (LC-DVC) [2]. Our proposed codec architecture can support real time processing @15fps. Keyword; distributed Video Coding (DVC), dynamic GOP control, classification, video surveillance, background modeling, codec structure 2 Shuai Li, Joohee Kim, Ken Choi