摘要： 传统的非标定图像校正都是依赖于基础矩阵的精度。而该方法根据图像校正基本准则，提出一种无需高精度基础矩阵的鲁棒性图像对校正算法，基础矩阵只是提供迭代优化的初始数值，并且与需要基础矩阵的方法相比较，结果表明该方法有效地消除垂直视差，能够较好地处理射影畸变问题。

关键词: 图像校正；基础矩阵；射影畸变

Abstract: Traditional uncalibrated image rectification method all depends on high-accuracy fundamental matrix, while this new method proposesa new robust rectification algorithm of uncalibrated stereo pairs according to image rectification principle and the fundamental matrix only is initial value for iterative optimizing. By the experiment of t he new method compared to the other that depends on fundamental matrix, it shows that the method can effectively eliminate vertical disparity and deal with projective distortion very well.

Key words: Image rectification; Fundamental matrix; Projective distortion