적응적 트라이맵을 사용한 합성 영상 편집 시스템

Abstract

Commercial video blending technologies used in movies, broadcasts, and commercials are complex to utilize by ordinary users because they require expertise. In this paper, we propose an image regeneration system that can get similar image composition results with fewer operations than conventional commercial programs. The proposed system consists of three steps with a video classifier, an image search extractor, and a video blender. Composite image generator is implemented with emphasis on object, face, and background compositing functions. This system is composed of a Poisson Image Matting technique for natural boundary processing of images, a Haar-like feature technique for automatically tracking face positions, sizes and angles, and a Watershed Segmentation technique for easy separation between foreground and background. Especially, we propose a video editor with new blending technique based on an adaptive Tri-map, in order to improve the quality of blended video under various image conditions. The proposed system makes it possible to regenerate high quality blending images through simple manipulation of videos and images synthesis so that it can be easily used by unskilled public