Adaptive motion estimation technique for motion compensated interframe interpolation

Abstract

In this paper, we propose a new fast motion compensated interframe interpolation algorithm. First, we propose a fast algorithm that gets backward motion vectors by using forward motion vectors received from the transmitter. It is important to predict well both covered and uncovered regions in order to interpolate skipped frames at the receiver, therefore, backward motion vectors are needed as well as forward motion vectors. But, in the case of the conventional video communication systems, we usually receive the forward estimated motion vectors from the transmitter. To estimate backward motion vectors used to predict the uncovered region, the conventional interframe interpolation technique requires a large amount of calculation at the receiver because it uses exhaustive motion search such as full search block matching algorithm (FSBMA). In this paper, we reduce the computational complexity to get backward motion vectors by using the dynamic search range varied to forward motion vectors received from the transmitter. Second, we also propose the algorithm to obtain the motion vectors which are closer to the true object motion by utilizing the local characteristics of vector field. This algorithm makes it possible showing that the resulting motion vectors are more adaptive to the real object motion ever the conventional one. According to the simulation results, the proposed algorithm is better than the conventional one in subjective quality as well as in objective quality.