Edge-Based and Efficient Chromaticity Spatio-spectral Models for Color Constancy

Fast and accurate estimation of the transformation imposed by the illuminant to the colors of an image taken under that illuminant is of crucial importance in real-time computational color constancy applications. To this end, we present an edge based and an efficient chromaticity spatiospectral model which are modified versions of the spatiospectral method introduced by Chakrabarti et al [1]. As compared with the conventional color constancy methods, the spatio-spectral model improves the accuracy of estimation at the cost of increasing the execution time and storage dramatically. This increase makes the spatio-spectral model impractical and inappropriate for real-time applications. Our proposed methods aim at reducing the computational burden and required storage for the spatio-spectral modeling while retaining its accuracy of estimation. Evaluation of the performance of the proposed methods on a synthetic color image database and also the “Color Checker” database [2] are presented.