Least square filtering for low-delay transform domain Wyner-Ziv video coding

Blocking artifact is one of the annoying problems that can significantly decrease the objective and subjective quality of low bit-rate video. In the transform domain Wyner-Ziv (TDWZ) video coding, two operations could result in blocking artifacts, namely the Discrete Cosine Transform (DCT) based operation and the block based motion compensation. We introduce Least Square prediction (LSP) into TDWZ to reduce blocking artifacts. First we propose using Least Square Filter (LSF) as a bridge to integrate LSP into side information (SI) generation. Then we use LSF as a post-processing filter to TDWZ to further reduce block artifacts. Experiments show that the qualities of WZ videos are consistently enhanced by 0.3dB to 0.7dB after LSF.