摘要
匹配是立体视觉中一个重要问题,它通常是在图像极线校正的基础上实现的.常用的校正方法会造成投影变换后图像拉伸不均匀,从而影响后续模板匹配的精度.为此提出一种基于极线校正逆变换(反校正)的立体匹配方法,它以校正后的图像进行开窗搜索,但针对窗内的图像坐标进行反校正,提取校正前的像素灰度值,最后进行模板匹配.仿真和实验结果证明,文中提出算法的匹配效果比较理想,生成的视差图要比传统算法精确.
Matching is important in stereo vision,it is often realized on the basis of epipolar line rectification.The common rectification method can affect the accuracy of template matching because of the asymmetry of image stretch after projective transformation.A stereo matching algorithm based on invert transformation of epipolar line rectification（invert rectification） is put forward in this paper,it completes the template matching based on pixel gray value of image coordinate in search window after invert rectification.The results of simulation and experiment show that the matching effect is desirable,the disparity map generated by our algorithm is better than traditional algorithm.
出处
《沈阳理工大学学报》
CAS
2010年第2期70-73,共4页
Journal of Shenyang Ligong University
基金
辽宁省教育厅基金项目资助(20080611)
关键词
极线校正
逆变换
立体匹配
image rectification
invert transformation
stereo matching
