一种稳健的图割立体匹配方法
A Robust Method Based on Graph Cuts for Stereo Correspondence
-
摘要: 为了提高传统的基于图割的立体匹配算法的鲁棒性, 提出一种基于图像增强的图割立体匹配方法, 传统算法是在图像灰度值的基础上构建能量函数, 该方法加入了图像的梯度值来构建能量函数, 然后将基于二值化标号函数的α扩展算法和KV最大流算法结合起来求解能量函数.由于梯度对于图像中的噪声、局部光照具有鲁棒性, 因此该算法在一定程度上增加了传统GC算法的鲁棒性, 最后通过仿真和实验验证了算法的有效性.Abstract: To improve the robustness of the algorithm based on graph cuts, a new method based on graph cuts is introduced in this paper. We formulated the energy functions based on gradient. Then the method was applied to minimize the energy function. As for the robustness of the gradient for the image noise and local illumination, the method increase the robustness of the traditional algorithm based on graph cuts. At last, the effectiveness of the algorithm is verified by simulation and implementation.