基于汇聚立体视觉的图像校正算法

通过分析汇聚立体视觉系统立体图像间的几何关系，提出了一种简单快速的图像校正算法，该算法利用矩阵的ＬＵ分解法，直接求解校正的转移关系矩阵．将立体视觉中对极几何的匹配关系变为图像扫描线的一一对应关系，从而将沿对极线搜索匹配点的过程变为沿图像行搜索的过程．实验表明该算法简单有效，具有较强的实用价值．

The rectifying algorithm for images based on convergent stereovision

By analyzing the relationship between the stereo image pairs, this paper presents a simple rectifying method, that is, the rectifying matrix that could be calculated using the matrix LU decomposition algorithm. It enables the matching constraint in epipolar geometry to be changed into the scan lines in image pairs, and the search for the corresponding image points to be confined to one dimension (that is, the epipolar lines are parallel to image scan lines after rectification). The experimental results show that the method is simple and practicable.