Color restoration of lighting scenes with locally adapted HDR images

Images of scenes illuminated by colored lightings are often degraded owing to the clipped whites and crushed shadows caused by the lack of dynamic range. Such degradation can be improved by introducing high dynamic range (HDR) images that can fully capture the brightness and color at a large bit-depth. Color appearance model (CAM) is widely used for accurately restoring colors, by replicating the property of human vision that can adaptively recognize colors according to the brightness within a local region. This local adaptation model, however, is inapplicable to lighting scenes where the spatial variation of brightness becomes large. We propose a method of combining such local adaptations separately applied for different regions on a single image. This method tries to restore the color of HDR images while retaining the global consistency of brightness variations. The effectiveness of our method is experimentally demonstrated for the indoor scene illuminated by a color LED.