Closed-form stereo image rectification

Abstract

Stereo image rectification transforms a pair of images into a new pair such that the epipolar lines in the transformed images have the same direction as the image rows and matching epipolar lines in different images have the same row index to enable an efficient and reliable dense stereo matching. In this paper we propose a first closed-form algorithm to automatically rectify stereo images just using the fundamental matrix. The algorithm involves only direct and purely geometric transformation processes. There is no iterative or optimization processes for the rectification parameter estimation in our method. Real images have been used for testing purposes, and convincing results have been obtained from our algorithm. Comparison with the state-of-the-art method shows that our method produces better rectification results.