Coding of multiview imagery with motion and disparity compensated orthogonal transforms

We consider the problem of multiview video compression with recently introduced motion-compensated orthogonal video transforms. It is well known that so called motion-compensated lifted wavelets may substantially deviate from orthonormality due to motion compensation, even if based on an orthogonal or near-orthogonal wavelet. Therefore, inaccurate motion compensation may result in a motion-dependent suboptimal subband decomposition. On the other hand, motion-compensated orthogonal video transforms are designed such that orthonormality is strictly maintained for any motion field. Even inaccurate motion compensation leads to an orthogonal subband decomposition. This is of particular interest for complex motion and disparity fields. For subband coding of multiview video, motion-compensated subbands will be further processed by disparity-compensated subband decompositions. In case of suboptimal subband decompositions, this cascade of processing steps may result in further inefficiencies. On the other hand, motion or disparity compensated orthogonal transforms offer the advantage that the subbands are always orthogonal, even if motion or disparity compensation is inaccurate. This paper investigates the energy compaction of motion and disparity compensated orthogonal transforms.