A 3D Lifting Based Method Augmented by Motion Compensation for Video Coding

Abstract

This study introduces a spatio-temporal lifting based algorithm to be used in compression of video signals. The temporal correlation of consecutive frames causes temporal redundancies, which are subject to lifting-like motion predictive compression. Similarly, neighbouring pixels are correlated within each frame. A method that uses both correlations might be 3D lifting-based decomposition. In this study, block-based motion compensation is added to the classical 3D lifting method. Domain of motion compensation is first selected as free, and then reverse-symmetric. It is observed that reverse-symmetric motion compensation improves the performance of the prediction step in 3D lifting based coding.