图像的区域立体匹配是立体视觉中的重点研究课题之一,实时可靠应用的关键在于视差图的可靠性和计算复杂度。提出了一种基于线性生长的区域立体匹配算法,实现从立体图像对中提取深度信息,获得更可靠视差图的方法。该算法包括根点选择和区域生长2个部分,获得视差图的计算时间短,利用滤波可以提高视差图的可靠度。最后对此算法生成的结果进行了比较分析。
Image stereo matching is one of the most important issues in stereo vision.Reliability of depth maps and computational cost of algorithm is the key point for implementing real time robust applications.In the paper,a region based stereo matching algorithm is developed for extracting depth information from a color stereo image pair,which is based on region growing.The algorithm includes two parts,i.e.root selection process and region growing process,and needs less calculation time to obtain disparity map.Filtering is adopted to increase the reliability of disparity map.Finally,the obtained results using this algorithm are presented and compared.