Variable Block Size Prediction For Wyner-Ziv Video Coding Using Structure Similarity

This work focuses on generating better side information by exploiting the gains of variable block size based motion-compensated interpolation (VBMCI). Wyner-Ziv video coding (WZVC) conduct temporal prediction at the decoder, and this leads two results: temporal prediction has no access to the current frame and there is no motion vectors sent from the encoder to the decoder. Both make traditional variable block size prediction techniques not suitable for WZVC. In this paper a new algorithm VBMCI-SSIM is proposed, in which the structure similarity is firstly used to aid controlling block splitting and finding correct motion trajectory. Experimental results prove the validity of the proposed algorithm.