Color Appearance in Multispectral Radiosity L. Neumann, 1 , 2 F. Castro 1 A. Neumann 3 and M. Sbert 1 1 Institut d’Informàtica i Aplicacions. Grup de Gràfics de Girona (GGG). Universitat de Girona. Girona (Spain) 2 Institució Catalana de Recerca i Estudis Avançats (ICREA) Barcelona (Spain) 3 Institut fur ComputerGraphik und Algorithmen. TU Wien. Wien (Austria) Abstract In closed environments, especially in bright colored interiors, there occurs a significant change of saturation and some shifting of hue of originally selected colors. This is due to multiple light inter-reflections. The human vision mechanism partly reduces this effect thanks to the change of the reference white. We use in this paper a multispectral radiosity method to describe and compute the physical effects. We also use a color appearance model, the new and powerful CIECAM02 model, to compute the perceptual aspects. The CIECAM02 includes the luminance and chromatic adaptation effects, and it has compact forward and inverse transformation formulas. The input data for the color appearance model is ensured by computing the radiosity so- lution, thereby there are known both the spectral radiance for every viewpoint and view direction and the spectral irradiance on every patch of the scene. Nearly all of earlier global illumination approaches ignored the often strong changes of originally selected colors. This paper supports the color environment design. Using the presented method it is possible the selection or mixture of paints to achieve, after the physical and perceptual effects, a color appearance previously selected under standard viewing conditions in a color atlas. 1. Introduction The color appearance problem has two components: the physical effects and the perceptual aspects. The physical model of the light interactions is described by the render- ing equation 16, 6 in general environments and by the radios- ity equation 4, 5 in diffuse environments. The more widely used approach uses only 3 color channels, that are generally enough for realistic rendering. However, an accurate description of colors requires a mul- tispectral approach 10 . These techniques use discrete spectral lines or a step-wise constant approach on the visible spec- trum for all components of the model (from reflectivity to light sources). In a radiosity approach, the number of val- ues needed for each patch is the number of these spectral intervals. On the other hand, the full spectrum models de- scribe a lot of interesting phenomena missing in a simple RGB approach. There appear new characteristics, like e.g. the metamerism or the possibility of exact photometric lu- minance calculation based on spectral luminous efficiency of human eye. After inter-reflection computation, the results can be sig- nificantly different between a simple RGB model and a mul- tispectral one having the same starting tristimulus color co- ordinates (e.g. CIE XYZ) for all surfaces and light sources in the scene 10, 11 . A drawback of the multispectral approach, especially in case of fine spectral resolution, is the high computational cost. This problem can be overcome by means of parallel computing. Another practical problem of the multispectral approach is often the lack of spectral data. We deal in this paper with diffuse scenes (because of sim- plicity and speed), but the extension to non-diffuse environ- ments, using one of the existing Monte Carlo methods 22 (like ray-tracing, Metropolis,etc.), is very easy. The novelty of the paper is not a new global illumination technique, but the combination of the multispectral global illumination with an accurate color appearance model to describe the color changes and to select the appropriate paint to invert or com- pensate the color changes. To illustrate with a very simple case the color shifting let