Abstract: A new stereo matching approach based on matching primitives from improved rank transform about color image is proposed.This algorithm extends the rank transform ideas and proposes a new rank transform equation by color similarity measures with absolute-value exponent method to solve the problems such as image noise and brightness difference in stereo images.Then，by adding chromatic gradient constrain，disparity image is done by a color image which is transformed into improved rank image.The experimental results show that this algorithm can obtain more precise disparity map，which demonstrates the practical value of the algorithm.

Key words: stereo matching, rank transform, improved rank transform, chromatic gradient

摘要： 将非参数Rank变换引用到彩色图像的立体配算法中，采用绝对值指数方法来计算像素的颜色相似性，对彩色图像进行一种分等级的改进Rank变换，将彩色图像转变了一个整数矩阵，然后建立一种新的匹配测度，同时加入色差梯度约束进行匹配。通过对比实验，表明这种新的匹配基元进行匹配，与直接采用灰度特征作为匹配基元的算法相比，有更好的抗噪性和匹配准确性。

关键词: 立体匹配, rank变换, 改进rank变换, 色差梯度