Low complexity image interpolation method based on path selection

摘要

Image interpolation, the problem of producing a sequence of intermediate frames between two input images, is of significant interest. The application of image interpolation is numerous, including animation of still images, view interpolation and temporal interpolation. In this paper, we achieve an improved path-based interpolation method based on the original path-based framework, by introducing two innovative improvements. On one hand, we use optical flow to decide the direction of path, to constraint the path length and to maintain the global path coherency, which improves the efficiency significantly. On the other hand, we introduce the pixel interlacing model to obtain more accurate optical flows so that the accuracy of path selection will be improved a lot. Our improved path-based method performs as well as the original method in various interpolation applications in image quality and surpass the original method by a large scale in efficiency.