Thresholding hierarchical block matching algorithm for motion estimation

The well-known full search block matching method is very time-consuming as a result of exhaustive search where all the possible motion vectors are considered. Multiresolution block matching algorithms reduce the computation by taking advantage of pyramid structure. However, both approaches treat every region in an image domain indiscriminately no matter whether the region under consideration contains complicated motion or not. In this paper, we present a new hierarchical block matching algorithm for motion estimation using thresholding. Regions experiencing relatively uniform motion are withheld from further processing via thresholding, thus saving lots of computation. No pyramid is actually formed, that is, all levels in the hierarchy have the same resolution. Experiments have demonstrated that for videoconferencing sequences, the proposed algorithm works faster and has much less motion vectors than both the full search block matching and multiresolution technique while maintaining almost the same quality of reconstructed images.