一种基于区域颜色信息的宽基线立体匹配算法

立体匹配技术是立体视觉研究领域的难点,利用图像的特征信息解决对应问题是研究的热点.本文在提取基于边缘仿射不变区域检测器的基础上,设计了基于区域颜色信息和几何信息的彩色图像立体匹配算法.该算法首先对所提取的不变区域进行颜色空间量化,在此基础上根据两种色彩之间的距离进行色彩聚类,然后计算直方图的二次距离,并进行相似性度量,最后根据区域间几何性质进行误匹配剔除.实验证明,该算法能降低解决对应问题的复杂性,提高正确匹配率.该方法对于目标识别、三维重建、非接触测量等研究均具有重要的意义.

A Wide Baseline Stereo Matching Algorithm Based on Color Information

The stereo matching technology is the difficulty in the field of stereo vision,and using the features of image to solve the correspondence problem is the focus of research.The color image stereo matching algorithm based on the region color information and geometry information is designed.Its basic method is detecting the affine invariant regions based on edge firstly;then quantizing the color space of regions detected,clustering the color by the distance between two colors,and computing the square distance of color histogram to measure the similarity of the correspondence regions;finally,eliminating the error matching region by geometrical information of regions.The experiment result show the algorithm can reduce the complexity to solve the corresponding,and improve the correct match rate.The method is important to object recognition,3D reconstruction and non-contact measurement research.