Spectral adaptation with uncertainty using matching

The authors use colour as a cue to estimate the presence of target objects in an image, by making assumptions about the spectral properties of surfaces in the image. They compute a match with the colour knowledge database using a least mean square error measure. A by-product of this match is a diagonal matrix spectral transformation which enables the system to adapt to varying illumination conditions. The fitness measure can be combined with other algorithms (e.g. edge detection, colour segmentation) and applied to the image to confirm the identity and the location of the objects. They present the implementation details of this simplified colour constancy algorithm, and provides results of its application to a number of test images