Projective Rectification with Minimal Geometric Distortion

Abstract

This chapter presented a new way of parameterizing the homography, which leads to a new approach of projective rectification for stereo images. Compared with the previous works, the novelty of this new algorithm is that it uses mean-square distance as minimization criterion which has more well-defined geometric meaning. Furthermore, instead of putting constraint on x-axis disparity, we use shearing transform to achieve a single solution for the projective rectification problem, and greatly reduce the geometric distortion. Visual inspection and quantitative evaluation of the rectification results show the accuracies of the proposed method and its low geometric distortion. Experiments on different types of image pairs with various y-disparity values have been conducted, and the results show that the proposed method can effectively reduce the geometric distortion.