The Ratio Method for Multi-view Color Constancy

Color constancy is the ability to infer stable material colors despite changes in lighting, and it is typically addressed computationally using a single image as input. In many recognition and retrieval applications, we have access to image sets that contain multiple views of the same object in different environments; we show in this technical report and a related publication [8], that correspondences between these images provide important constraints that can improve color constancy. In this report, we present another method to solve the multi-view color constancy problem, the Ratio Method. This method provides a means to recover estimates of underlying surface reflectance based on joint estimation of these surface properties and the illuminants present in multiple images. In contrast to the multiview Spatial Correlations method (MVSC), this method can leverage any single image color constancy method as a bootstrap for the multi-view solution. The method exploits image correspondences obtained by various alignment techniques, and we show examples based on matching local region features. Our results show that the Ratio Method performs similarly to the MVSC method, both of which are improvements over a baseline single-view method.