In order to improve red-green color blindness' ability of distinguishing colors, this paper introduces a partial rectification of color blindness based on image-segmentation. An emulation method of color blindness is studied first; then the original image is segmented by K-means cluster algorithm and Hierarchical cluster algorithm. Then, it calculates Euclidean distances of each region in LAB Color Space as the metric of color similarity, and confirms which color regions are hard to distinguish by red-green colorblindness. Finally, the region is replaced by other color with the same brightness and a higher degree of differentiation. With the test of the algorithm on Matlab platform, the result shows that this method can improve color blindness' ability to distinguish colors and it performs better than other existing methods in reducing cognitive deviation of color.