Low complexity multiple candidate motion estimation based on constrained one-bit transforms

Abstract

In this paper, we propose a low complexity multiple candidate motion estimation algorithm based on the constrained one-bit transform. We propose variations of constrained one-bit transform whose matching criteria are almost the same as the constrained one-bit transform. The motion estimation performances of the proposed variations are statistically similar to that of constrained one-bit transform in whole, but its local behaviors are very different. By adopting the multiple candidate search strategy into the typical constrained one-bit transform and its variation thereafter, we can efficiently determine two best motion vectors and enhance the overall motion estimation accuracy. Experimental results show that the proposed algorithm achieves peak-to-peak signal-to-noise ratio gains up to 0.66dB on average compared with the conventional constrained one-bit transform-based motion estimation with negligible complexity increase.