An Iterative Approach for Color Constancy

The color perception of a surface as invariant under changing illumination, called color constancy, is very important in applications such as outdoor navigation. Most color constancy approaches are targeted for single images obtained from cameras whose sensors are assumed to be narrow-band. Algorithms that take advantage of the related information in consecutive frames are rare and require the presence of known colors in each frame or unrealistic assumptions. Therefore, in this work, we propose an iterative method for color constancy, in which the reference colors estimated from images previously corrected are used for correcting the colors of the new images through a color mapping matrix. This matrix achieves better results compared to a diagonal matrix when sensors are not narrow-band. The iterative method results are very similar to frame by frame correction using Color Checker standard colors values.