Block-recursive matching algorithm (BRMA) for displacement estimation of viedo images

This work proposes a block-recursive matching algorithm (BRMA), for displacement estimation. Simulation results show that three iterations are adequate to almost reach the asymptotic performance and the improvement of the recursive method versus the nonrecursive method varies significantly from frame to frame. A smaller block size was shown to result in better performance. However, it requires more overhead bits to send the motion vectors and temporal coefficients. In an mnterframe motion-compensated hybrid system with transform coding for intraframe coding and differential coding for temporal coding, a larger block size always leads to better performance. On the other hand, a larger block size always results in a poorer motion compensation. To obtain the advantages of large block sizes in transform coding and small sizes in motion compensation, we suggest an mnterframe motion-compensated system with a small block size in the displacement estimation step and a large block size in the transform coding step.