基于mean-shift全局立体匹配方法

针对图像全局立体匹配精度高、计算量大的问题,提出基于mean shift图像分割的全局立体匹配方法。首先,通过mean shift算法对图像进行分割,获取图像同质区域数量和区域的标号。在计算匹配代价时,根据像素所属的分割区域,对像素进行筛选,从而提高匹配代价计算速度;其次,在代价聚合前,将mean shift算法获取的同质区域数K值赋值给K-means聚类算法,对像素再次聚类,提高立体匹配精度和速度;最后通过TRW-S置信传播解决能量最小化问题。实验表明,该算法明显提高了匹配的准确性和速度,与单纯的全局匹配算法相比,具有更大的优势。

A global stereo matching algorithm based on mean-shift

We propose a global stereo matching algorithm based on mean shift image segmentation to improve image global stereo matching for its high accuracy but large calculation. Firstly, we use the mean shift algorithm to segment the original image to get the number of homogeneous regions and their labels. When calculating matching cost, we choose proper pixels according to the segmentation region of pixels, which can improve the computation speed of matching cost. Secondly, before calculating the cost aggregation, we use the K-means algorithm to cluster the pixels according to the number of homogeneous regions K which is obtained by the mean shift algorithm before. This can improve the accuracy and speed of stereo matching. Finally, we utilize the TRW-S belief propagation algorithm to solve the energy minimization problem. Experimental results show that compared with pure global stereo matching, the proposed algorithm can improve the stereo matching accuracy and speed obviously.