Abstract
Cross-dissolve technique is one of the typical video effects for the scene change. The existing video coding standard such as H.264/AVC and H.265/HEVC does not work well especially for the video effect scene. The reason is that such video scene may include two types of object motion corresponding to fade-in and fade-out video, and the mixed motion degrades the performance of motion compensated prediction. In order to improve the coding performance of the cross-dissolve scene, we propose a weighted prediction method that can predict fade-in video characteristics even while encoding fade-out pictures in order to improve the weighted prediction performance.
