Estimation Illumination Chromaticity

Color constancy is important for pattern recognition and image understanding. Based on simplified BRDF model, we use voting method to detect the high light areas. Then the pixels in the high light areas are projected to the inverse-intensity space. In the inverse-intensity space, the correlation of image chromaticity, inverse-intensity, and illumination chromaticity is linear. So we can estimate the Illumination chromaticity. The experiments demonstrate that our algorithm have no limit on the objects´ material and texture and contain the color constancy after the illumination chromaticity correction.