Abstract: A new area-based stereo matching approach in rank transform domain is proposed in this paper.First，stereo image pairs are mapped from the gray-level space to rank space to eliminate the interference due to the environment variations and dissimilar characteristics of different cameras.Based on the assumption that most objects interested in a picture have similar depths of fields，a two-level stereo matching is processed in the rank domain.The global disparity is first estimated in the maximum window and then the relative disparity is estimated.The final disparity is the sum of global disparity and relative disparity.The experimental results indicate that the proposed algorithm can effectively reduce the error matching rate and is better than the classical area-based stereo matching algorithm.

摘要： 提出了一种基于秩空间的区域立体匹配算法。首先对立体图像进行秩（rank）变换，将图像从灰度空间变换到秩空间，消除因噪声和两个摄像机参数不一致产生的干扰，再根据自然视频序列每幅图像多数景物景深变化不大的事实，把视差分解为全局视差与局部视差之和，在秩空间进行二次立体匹配：先在最大窗口内估计全局视差，然后在这个最大窗口内采用块匹配方式进行二次匹配求各点的实际视差。该二次立体匹配算法有效地消除了误匹配，提高正确匹配率。实验结果证明，提出的算法优于传统的基于区域的立体匹配方法。