Robust Algorithm for Computational Color Constancy

Color constancy (CC) or white balance (WB) is an important image processing step in digital camera and machine vision. This paper presents a robust computational algorithm that offers an accurate solution for CC. The proposed CC measure exploits the scanning strategy of RGB channel gain adjustment. In addition, the suppressing mechanism of grayscale pixel maximization (GPM) is used to locate potential illuminants. The numerical analysis scheme provides GPM diagrams for user-friendly visualization and analysis. For most natural images having a noticeable amount of grayscale pixels in the scene, the channel gain adjustment required for CC can be clearly revealed. Operating mechanism and CC performance of the proposed algorithm are intelligibly demonstrated.