Motion Estimation Using Spatio-Temporal Contextual Information

Abstract

Motion estimation is an important and computationally intensive task in video coding applications. Block-matching-based fast algorithms reduce the computational complexity of motion estimation at the expense of accuracy. An analysis of computation and performance trade-offs in motion estimation algorithms helps in choosing a suitable algorithm for video/visual communication applications. Fast motion estimation algorithms often assume a monotonic error surface in order to speed up the computations involved in motion estimation. The argument against this assumption is that the search might be trapped in local minima resulting in inaccurate motion estimates. The paper investigates state-of-the-art techniques for block-based motion estimation and presents an approach to improving the performance of block-based motion estimation algorithms. Specifically, the paper suggests a simple scheme that includes spatio-temporal neighborhood information for obtaining better estimates of the motion vectors. The paper also investigates the effects of the monotonic error surface assumption and suggests a remedy to reduce its impact on the motion field estimates. The presented experiments demonstrate the efficiency of the proposed approach.