Color Management within a Spectral Image Visualization Tool

Recent developments in spectral imaging are pointing toward a future where the demands on color management will require a richer infrastructure than that which is currently offered. ICC color management includes a stage where all colors are transformed to and from XYZ-based colorspaces. This colorimetric bottleneck is acceptable within a metameric or Maxwellian approach to color reproduction, but severely undermines the advantages of spectral imaging. Spectralizer, a spectral image visualization tool, has been implemented to provide a platform where spectral images may be easily displayed, manipulated, analyzed and processed. It has proven to be useful in investigating algorithms and prototype datastructures for performing the management of color within a spectral imaging environment. Spectral Profiling of an Imaging Device Source Devices The spectral parameters to image capture are reasonably few. Figure 1 shows a simple schematic. A light source, or a number of light sources, radiate light illuminating objects or filtered through transparent or translucent objects. Some amount of the reflected or transmitted light travels through the optical system of an imaging device. As the radiation is followed, it encounters further filtration and then the detector itself. Light sources can be described by spectral power distributions, objects by spectral reflectance or transmittance properties, optical components and filters are associated with spectral transmittances and detectors have spectral sensitivities. If all of these properties are known and if other types of indirect illumination can be calculated from the scene then it is possible to create a model that reports the imaging system output given a particular input scene. As in all profiling tasks, the inverse of the imaging model is of interest. If one were interested in determining object spectral reflectance given a particular digital output, such models have been reported. However, without strict control over illumination, it may be impossible to create an inverse model that delivers reflectance for a given system response. When control over scene illumination cannot be guaranteed, an inverse model that reports scene radiance for a given digital output is feasible. Accuracy expectations for either model will depend on the character of system channels and the level of a priori knowledge of scene object characteristics. For scanning devices, full knowledge of scene illumination is not an unreasonable demand and thus a spectral profile for such a device could indicate reflectance for given digital output. However, for a camera likely to be used in arbitrary environments, profiling the relationship between system digital count and estimated scene radiance is the appropriate strategy.