Abstract: A novel stereo matching algorithm based on multi-window in rank transform domain for color image was proposed in this paper.It could overcome the interference of the environment variations and binocular cameras and reduce the matching error rate in depth discontinuous region.Firstly,we transformed color image from the standard image database into an integer matrix with improved rank transform.So the image was transformed from color space to rank space.Secondly,a new method calculated the color similarity with improved absolute-value exponent method.And the problems such as image noise and light difference in stereo images were solved.At last,disparity was estimated by multi-window algorithm with fast searching strategy.The experimental results showed that this algorithm could obtain robust matching results with strong anti-noise strong feature in real time.Meanwhile,it achieved an improvement for accuracy at 18.5% in depth discontinuous areas.