The image quality was degraded because of imaging in cloud and fog. The essence was atmosperic scattering for light between scene with observation point. The restoration method was projected for the images imaged under bad weather. First, the mechanism of atmosperic scattering was analyzed. Second, the attenuation model, airlight model and phase distribution function in directions of atmospheric scattering were introduced. Third, the formula of atmosperic light intensity was derived. Based on the models of attenuation and airlight and the pathlength was considered, the advanced multi-scale Retinex algorithm (AMSR) was applied for removing cloud in color remote sensing images. The AMSR method included three steps. The first one was to obtain the complementary image. The second one was that the complementary image was enhanced. The third one was to the complementary image from the enhanced image. Because of the sence information was lost following with removing cloud, the criterion of information entropy was applied to judge the quality of enhanced images. The results show that the AMSR was suitable for removing cloud in color remote sensing images. The best image was stretched in the range of k=1.5±0.1 times standard deviation near the mean brightness and the information entropy was largest.