Color constancy using achromatic surface

Abstract Although a number of elaborate color constancy algorithms have been proposed, methods such as Grey World and Max‐RGB are still widely used because of their low computational costs. The Grey World algorithm is based on the grey world assumption: the average reflectance in a scene is achromatic. But this assumption cannot be always satisfied well. Borrowing on some of the strengths and simplicity of the Grey World algorithm, W. Xiong et al. proposed an advanced illumination estimation method, named Grey Surface Identification (GSI), which identifies those grey surfaces no matter what the light color is and averages them in RGB space. However, this method is camera‐dependent, so it cannot be applied on the images from unknown imaging device. Motivated by the paradigm of the GSI, we present a novel iteration method to identify achromatic surface for illumination estimation. Furthermore, the local Grey Edge method is introduced to optimize the initial condition of the iteration so as to improve the accuracy of the proposed algorithm. The experiment results on different image datasets show that our algorithm is effective and outperforms some current state‐of‐the‐art color constancy algorithms. © 2010 Wiley Periodicals, Inc. Col Res Appl, 2010