摘要： 提出一种在图像分割基础上生成稠密视差图的立体匹配算法，利用滑动窗口生成鲁棒性较好的视差空间图DSI，运用分水岭算法将图像分割成多个小区域，根据每个区域在视差空间图内的视差分布情况计算该区域的信赖度，并综合各区域的邻域对其产生的影响，以获得稠密视差图，采用国际标准数据对该算法进行测试，实验结果表明，该算法能够获得较好的匹配结果。

关键词: 立体匹配, 立体视觉, 邻域相关性

Abstract: A novel stereo matching algorithm via generating dense disparity map based on image segment is proposed. A robust dense disparity-space image is generated by using sliding window. The watershed algorithm is used to segment the image into several small areas. According to the distribution of dense disparity for each area, the reliability of areas is calculated. By taking the impact caused by the neighbor areas into account, the calculation result is used to generate dense disparity map. The international standard data is used to test the performance of this algorithm. Experimental results show this algorithm can achieve better matching effect.

Key words: stereo matching, stereo vision, neighbor relativity

中图分类号: 

• TP301.6