A method for composition of real images and CG images based on feature points in images

Abstract

In this paper, we propose a system for composition of real images taken with a video camer and CG images. In general, we can consider that two problems arises when we try to generate high quality composite images. First problem is how exactly we acquire feature point's coordinate in images which are required to estimate geometrical relationship between real world and CG world. To deal with this problem, in a proposed system, we employ a method for acquisition of high resolution images which has been proposed earlier and acquire accurate coordinate of feature points with subpixel accuracy. Second problem is how easily we generate composite images with a video camera for home use. To this problem, we use a camera calibration method in which only four points on a plane surface are needed. At the end of this paper, we demonstrate that a composite image sequence which consists of 178 frames is successfully generated with a proposed method.