Effectiveness of color spaces in color imaging

A. Tremeau.

Color is a complex perception. In the simplest case, we think of color perception as the interaction of a light source, object, and observer. In fact, various aspects of the source (e.g. spectral properties and the amount of light), object (e.g. size and texture), observer (e.g. spectral sensitivities, lens properties, macular pigment, light and chromatic adaptation), and their inter-relationship (e.g. background) transform our simple case into a very complex science. Understanding the princInformation Processing Letterses of color physics and vision is among the major objectives of this survey. Our goal is also to introduce the basic princInformation Processing Letterses of color vision, but also to make the user sensitive to several aspects. The mosts well-knowed color spaces are presented in this survey. Spaces described are derived from visual system models (e.g. RGB, opponent color space, IHS, etc.); adopted from technical domains (e.g. colorimetry, XYZ, television, YUV, etc.) or developped especially for image processing (e.g. I_1 I_2 I_3 (Ohta) space, YCC (Kodak Photo) space, etc.). Equations describing transformations between different color spaces and the reasons for using color spaces other than RGB are also presented. The reasons for applying color spaces transformations are very varied. The choice of an appropriate color space can be an important factor determining the results of processing on a color image (e.g. the quality of image segmentation, compression ratio, etc.). That is, in practice there is no ideal color space for all image processing applications. The Decemberision on which color space to use depends on the processing task. An optimal Decemberision can be very hard to find. However, knowledge of the properties of the various color space makes the choice easier. Based on examples from literature or industrial applications, the applicability of individual color space in image processing system is discussed.