A Novel Content-Based Side-Information Generation Approach for Distributed Video Coding

The performance of Distributed Video Coding (DVC) has a great reliance on the accuracy of side-information, which is estimated according to the intra-coded key frames that are assumed to be reconstructed lossless in this paper. A novel content-based side-information generation approach is proposed to improve the Wyner-Ziv frame estimation. First, the motion features and texture features are extracted and used to segment the side-information into different regions with different confidence. Then the low-confident regionspsila side information could be corrected by the information obtained from the adjacent high-confident regions. At last, an adaptive block (AB) technique is implemented to improve performance on the base of the results from the content-based approach. Experiments illustrated that the average PSNR of side-information obtained by the content-based side-information approach with adaptive block is about 0.9 dB higher the cases without it.