Extraction of illumination effects from natural images with color transition model

Abstract

In this paper, we propose a method, which divides a natural color image into the surface characteristic of the object and the illumination effects. In proposal method, a spectral distribution of illumination source is approximated by a spectral distribution of black-body radiation. And we make the model of the change of chromaticity by the change of illumination condition. Using this color transition model, a natural image is divided into several regions by the surface characteristic and the clustering result of color space. And the proposal method judges whether the neighbor regions can be integrated. In addition, the entire image is divided by the regions which have a surface characteristic. A one standard color value in each region that shows the surface characteristic is decided with using color transition model. As a conclusion, we got a shadow-less image by the combination of the standard value of pixels and particular illumination effects. © 2008 Springer Berlin Heidelberg.