Distributed Video Coding with Multiple Side Information Sets

摘要

This letter proposes a method to retrieve the original image X out of multiple sets of SI (Side Information) in distributed video coding (DVC). Using Turbo decoding methods, the most reliable segments from the decoded Y_I's were selected for the composition of Y_∞, whose conditional entropy H(X∣Y_∞) became much lower than any individual conditional entropy H(X∣Y_I). This proposal has improved the peak signal-to-noise ratio (PSNR) by 1.1 to 1.8 dB, compared to the conventional single Si-based approach.