基于区域增长的立体像对稠密匹配算法

Image Dense Stereo Matching by Technique of Region Growing

该文提出了一种新的基于区域增长的立体像对稠密匹配算法 ,该算法适用于多种图像对 ,包括存在较大视差的未经校准的图像对和其中某些纹理稀疏的区域 .首先用新的两层算法匹配图像对中的种子点 ,匹配关系再根据两种策略由这些种子点向图像的其余部分传播 .区域增长过程中以新的加权差值平方和准则作为目标函数 ,模板窗的大小根据其中包含的纹理数量动态变化 ,而搜索窗的大小与可信系数成反比 .对真实立体像对的稠密匹配结果表明了该算法具有良好的性能 .

A new dense stereo matching algorithm is presented based on technique of region growing, which can be applied to a wide range of image pairs including those with large disparity or without rectification even if part of the images are less textured. Firstly, a number of feature points in the image pair are matched using a new two-level matching method. Then these reliably matched points are taken as the seeds from which corresponding relations propagate towards other regions of the images under two strategies. Weighted SSD (Sum of Squared Difference) is developed as the target function. The size of the template window is adaptive with textures within it and the size of the search window is varied in inverse proportion to the confidence coefficient. The algorithm has been tested with real stereo images and the results demonstrate its accuracy and efficiency.