Abstract
To transfer color images between color devices, we seek a representation that efficiently encodes the color information necessary for accurate reproduction. Our representations emphasize physical properties of illuminants and surfaces: the illuminant’s spectral power density and the surface’s spectral reflectance function. Encoding these spectral functions with finite-dimensional linear models facilitates analysis and computations based on them, but discards some of the information they contain. Existing techniques for computing linear models have emphasized accuracy in approximating the spectral functions themselves. For our applications, however, accurately approximating spectral functions is far less important than accurately approximating the receptor responses of a set of color devices, for example, a camera, scanner, or the human eye, to a set of surfaces and illuminants. Here we introduce a new technique for building linear models that takes a collection of devices, surfaces, and illuminants, and builds linear models that are optimal for predicting receptor responses.
© 1991 Optical Society of America
PDF ArticleMore Like This
David H. Brainard and Brian A. Wandell
WD4 OSA Annual Meeting (FIO) 1988
Brian A. Wandell
TuV2 OSA Annual Meeting (FIO) 1991
Hongyun Wang, Jing Wang, Guoxin Rong, and Björn M. Reinhard
FWL1 Frontiers in Optics (FiO) 2011