Study on motion estimation using local minima
Min-Sup Lee, Shiryong Yu · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 1993
Block matching algorithm has been widely employed in motion estimation, because of its simplicity of implementation. Among them the full search algorithm gives the optimal solution within the search area except the exhaustive computational cost. On the contrary, other fast search algorithms have the shortcomings of finding the local minimum due to coarse sampling positions. In this paper two issues, speed and local minimum, are dramatically improved by using the line projection and local minima. The proposed algorithm is constructed of two stages. First some candidates, local minima, for global minimum are selected by using the full search on the line projection data. And then these candidates are checked by general block matching algorithm. The performance of the proposed algorithm approaches that of the full search algorithm, but the computational cost is much less than the full search algorithm.