Video coding with 3D wavelet transforms

Abstract

Video coding systems based on 3D wavelet transforms offer several advantages over traditional hybrid video coders. This thesis proposes two 3D wavelet-based video-coding approaches. In the first approach, motion compensation with redundant-wavelet multihypothesis, in which multiple predictions that are diverse in transform phase contribute to a single motion estimate, is deployed into the fully scalable MC-EZBC video coder. The bidirectional motion-compensated temporal-filtering process of MC-EZBC is adapted to the redundant-wavelet domain, wherein transform redundancy is exploited to generate a phase-diverse multihypothesis prediction of the true temporal filtering. In the second approach, a video coder is proposed that does not perform motion compensation explicitly, instead relying on the motion-selective characteristics of the 3D dual-tree discrete wavelet transform to isolate moving features. The transform coefficients are coded with binary set-partitioning using k-d trees in an algorithm that exploits within-subband spatiotemporal coherency as well as cross-subband correlation to achieve efficient coding.