MOTION-COMPENSATED ORTHOGONAL TRANSFORMSFOR MULTIVIEW VIDEO CODING

摘要

This paper discusses coding of multiview video with motion-compensatedrnorthogonal transforms. These transforms have recentlyrnbeen introduced to generate strictly orthonormal decompositions ofrnimage sequences. Orthonormality is maintained for arbitrary motionrncompensation among the images. This is of particular interestrnfor coding applications as motion compensation is inaccurate duernto quantization of motion information. This is in contrast to thernwell known motion-compensated lifted wavelets where the propertiesrnof the decomposition are motion-dependent. This disadvantagernaffects compression efu0002ciency, and in particular that of multiviewrnvideo signals. Multiview video coding schemes utilize motion andrndisparity compensation when decomposing the multiview imageryrninto view-temporal subbands. Usually, decompositions in time andrnview direction are cascaded. Motion-compensated lifted waveletsrnsuffer from their motion-dependent decomposition, and in particularrnif cascaded for multiview video coding. On the other hand,rnmotion and disparity compensated orthogonal transforms offer thernadvantage that the overall decomposition is still strictly orthonormal.rnThis paper investigates the advantages of strict orthonormalityrnthat is offered by motion-compensated orthogonal transforms.