基于颜色与空间距离的置信度传播立体匹配算法

立体匹配是计算机视觉领域的一个重要研究课题,为了得到准确、稠密的视差图,提出了一种基于颜色与空间距离的置信度传播立体匹配算法.该算法首先将立体匹配表示成马尔科夫随机场模型,然后利用置信度传播方法来解决马尔科夫随机场中的最大后验估计问题,以达到建立稠密视差图的目的.算法中利用了彩色图像的颜色与空间信息建立视差空间图,并使用削减的线性模型来定义视差图的平滑项.采用了标准测试图像进行了测试,结果表明算法具有很好的性能.

Belief Propagation Stereo Matching Algorithm Based on Color and Space Distance

Stereo matching is one of the most important researches in computer vision. In order to get dense and correct disparity,a stereo matching algorithm based on color and space distance and using belief propagation was proposed. Firstly,Markov Random Field model was used to build the stereo matching model,and then belief propagation was considered to resolve the MAP(Maximum A Posterior) of MRF,so as to recover the dense depth map. Color and space information of color images was used to build the disparity space image,and the truncated linear model was used to define the smooth term of disparity. Some standard images were used to test the algorithm,and the experiment result shows that the algorithm is of good performance.