Stereo matching for 3D building reconstruction

Abstract

In this paper, we present an approach for a 3D building reconstruction from stereo images using local matching techniques. The approach starts with local intensity based stereo matching techniques which utilizes, sum of square differences (SSD) and gradient-based matching techniques. The obtained disparities (left and right disparity) using these matching techniques are filtered out by applying cross-checking algorithm i.e. by comparing left-to-right and right-to-left disparity maps for increasing the reliability of the disparity map. After the aggregation process, the winner-take-all optimization is used to find the optimal disparity map. Furthermore, in order to obtain better disparity, a median filter is adopted for preserving boundary of image and effective removal of noise. The results show that the proposed scheme is reliable, accurate and robust to high resolution aerial images. Results are also compared with the ground data. © 2011 Springer-Verlag Berlin Heidelberg.