Intermediate view synthesis based on disparity estimation and image interpolation

Abstract

Virtual view synthesis is the key technology of virtual reality and multi-view video system. Thus a new algorithm based on disparity estimation and interpolation is proposed in order to achieve intermediate view of any position between original images. A method of precise matching of every pixel based on block matching in the whole image is proposed in disparity estimation. And prediction work of start point of searching is also adopted to decrease computation of disparity estimation. Image interpolation is adopted to synthesize virtual view. Reverse mapping is used as an optimizing approach to fill up the holes formed in interpolation. Satisfying visual effect is shown in experimental results when there exist few sharp edges in the scene.