디포커스와 디스패리티 좌표 변화량의 관계 연구 및 활용을 통한 스테레오 매칭의 정확도 향상

Abstract

Integrating the shape from stereo and the shape from defocus/focus could enhance depth rendition. However, combining the two methods with different depth results is a very difficult problem. As an alternative, combining the disparity and blur before depth estimation may yield better rendition. In this context, this paper suggests a geometric relationship between the disparity and blur in rectified images. The derived relationship, called {\em differential stereo-defocus equation}, makes it possible to closely couple the stereo and blur in image planes.To check the differential stereo-defocus equation, we tested stereo matching which is constrained heavily by blur information. As can be expected from the equation, this approach can resolve the uncertainties in depth where the object surface has some slope over the image plane.