Color normalization for the robust recognition of television broadcasts

This paper presents a new procedure for the correction of color information in the context of the automatic recognition of video sequences on television. From its acquisition by a camera to its reproduction on a screen, lots of arti- ficial influences can lead to changes of the image. Since algorithms for the determination of visual correspondences usually work on basis of the color information of images, color distortions represent an important factor concerning the recognition accuracy. Purpose of the presented Color Gamut Sphering (CGS) procedure is to reduce distortions in the color information and therewith, to minimize the negative effect on the recognition reliability. It is a color normalization procedure, which works without any prior information on the content of video sequences and that is based on a color transformation that has proved to remove dependency on small brightness and saturation distortions. Its efficiency has been evaluated in comparison with wellknown color correction algorithms