Motion-compensated orthogonal transforms for multiview video coding

This paper discusses coding of multiview video with motion-compensated orthogonal transforms. These transforms have recently been introduced to generate strictly orthonormal decompositions of image sequences. Orthonormality is maintained for arbitrary motion compensation among the images. This is of particular interest for coding applications as motion compensation is inaccurate due to quantization of motion information. This is in contrast to the well known motion-compensated lifted wavelets where the properties of the decomposition are motion-dependent. This disadvantage affects compression efficiency and in particular that of multiview video signals. Multiview video coding schemes utilize motion and disparity compensation when decomposing the multiview imagery into view-temporal subbands. Usually decompositions in time and view direction are cascaded. Motion-compensated lifted wavelets suffer from their motion-dependent decomposition and in particular if cascaded for multiview video coding. On the other hand motion and disparity compensated orthogonal transforms offer the advantage that the overall decomposition is still strictly orthonormal. This paper investigates the advantages of strict orthonormality that is offered by motion-compensated orthogonal transforms.