부산대학교 도서관

This paper proposes an improved virtual view image synthesis algorithm to overcome the drawbacks of conventional virtual view synthesis technologies in dealing with holes and artifacts. First, conduct edge detection and preparation operation on obtained depth images to reduce the holes caused by the step change of pixel value; second, use the Euclidean distance between pixels and the depth information of image to roughly integrate images; finally, conduct morphological swelling treatment on the obtained virtual images to fill the rest holes and remove artifacts. According to results of simulation experiment, the improved algorithm outperforms those applied in other papers in filling holes and removing artifacts in in the course of virtual view image synthesis.