The Retinex model based on the theory of color constancy was introduced, and the principles and implement method of MSRCR were mainly analyzed. In order to verify that the algorithm based on Retinex theory performed well in the field of image enhancement, color road monitoring images collected in haze weather were regarded as experiment objects. MSRCR and traditional histogram equalization were respectively applied to process haze image in Matlab7.0. Subjective evaluation, image information entropy and luminance channel histogram were used to compare and analyze the results of these two algorithms in image enhancement. The results showed that the images processed by MSRCR restored richer details, higher image entropy than those of traditional histogram equalization and color of image after processing was closer to original one.