Object detail correspondence problem in stereovision

Abstract

The authors in their research deal with a broad spectrum of 3D reconstruction problems. In this paper, the authors explore the correspondence problem between some object details imaged with stereo vision head. Since the authors are working on a new matching algorithm based on graph theory, it was decided to test how problematic regions of images match each other. Problematic regions like edges, corners and other shapes with a shaded or motley background seem to create different images for left and right pictures and, in the case of further process, they may cause serious matching difficulties. Images of such problematic object details have been explored with several well-known edge detection methods, which have never been applied to stereovision before.