Abstract
Existing methods for white balancing photographs tend to rely on skilled interaction from the user, which is prohibitive for most amateur photographers. We propose a minimal interaction system for white balancing photographs that contain humans. Many of the pictures taken by amateur photographers fall into this category. Our system matches a user-selected patch of skin in a photograph to an entry in a skin reflectance function database. The estimate of the illuminant that emerges from the skin matching can be used to white balance the photograph, allowing users to compensate for biased illumination in an image with a single click. We compare the quality of our results to output from three other low-interaction methods, including commercial approaches such as Google Picasa's one-click relighting [19], a whitepoint-based algorithm [16], and Ebner's localized gray-world algorithm [7]. The comparisons indicate that our approach offers several advantages for amateur photographers.
