Analysis of spatial transform motion estimation with overlapped compensation and fractional-pixel accuracy

The authors investigate how the performance of spatial transforms motion es timation can be significantly improved by incorporating overlapped compensa tion and fractional pixel accuracy. An overlapped spatial transformation (O ST) motion model is developed, which successfully addresses the inability o f the conventional block matching technique to compensate for complex motio n and inside-block motion field discontinuities. Simulation results show th at the motion compensated prediction error of this method is reduced by 1.1 dB, compared with the conventional overlapped block matching motion estimat ion, for the same generated motion vectors overhead. To improve the perform ance, the overlap must be used in both the motion estimation and compensati on processes. Further improvement can be obtained using half-pixel precisio n motion vectors. However, this improvement is comparatively less than the gain of conventional block matching from a half-pixel search.