It is a difficult problem to choose appropriate image windows in area-based stereo matching algorithm. Most adaptive window algorithms utilize an evaluation function with disparity and intensity to choose the appropriate windows. This paper presents a new adaptive stereo matching algorithm based on edge detection. It chooses the appropriate image window only by the intensity variance and the process has no relation with disparities. This is the main difference between this algorithm and the others. Meanwhile we introduce a new rank transform method into the stereo algorithm. To demonstrate the effectiveness of the algorithm, it has been tested by stereo images with ground truth in Middlebury stereo database. The results show that our algorithm has high correct matching rate. Meanwhile it is robust and has few parameters to be set.