New adaptive interpolation schemes for efficient meshbased motion estimation

Motion estimation and compensation is an essential part of existing video codingsystems. the mesh-based motion estimation (mme) produces smoother motion field, bettersubjective quality (free from blocking artifacts), and higher peak signal-to-noise ratio(psnr) in many cases, especially at low bitrate video communications, compared to theconventional block matching algorithm (bma). however, the iterative refinement processof mme is computationally much costly and makes the method impractical in real- (or nearreal-) time systems. also, eliminating the iterative refinement step deteriorates the motionestimation result. in this paper, we propose motion adaptive interpolation schemes for noniterativemme, which use bma to compute the motion vectors (mvs) of mesh nodes. theproposed algorithm aims at compromising the mme and bma by modifying theinterpolation patterns (ipps) of the mme in an adaptive manner, based on the mvs ofmesh nodes. experimental results show notable rate-distortion improvement over bothbma and conventional non-iterative mme, with acceptable visual quality and systemcomplexity, especially when applied to sequences with medium to high motion activities.