Fast search algorithm for edge-oriented block matching algorithm

The edge-oriented block matching algorithm (EOMBA) provides a better motion-compensated prediction along moving edges in comparison with the traditional intensity-based block motion estimation methods. However, the accuracy of the motion vectors of blocks containing the boundary between the moving objects and the still area is critical in the EOBMA. Consequently, the full search algorithm (FSA) with a large search window is used to ensure its accuracy. In order to reduce the computational complexity of the motion estimation of these blocks, a fast search algorithm for the EOBMA is proposed in this paper. The proposed algorithm employs the edge feature to locate initial search points such that the computational burden of the motion estimation of these boundary blocks must be reduced. Results on applying the proposed algorithm to the EOBMA show that the proposed search strategy is able to strengthen the EOBMA.