There is an urgent need to build digital image databases with adequate colorimetric accuracy for museums, achieves and libraries. Traditional colorimetric imaging suffers from the possibilities of metameric problem, while spectral imaging can facilitate accurate tristimulus estimation and possibilities for spectral reconstruction of each pixel. Spectral image archives can be used to render accurate images both spectrally and colorimetrically to the original target for any illuminant and observer. The most convenient and practical capture system for spectral imaging combines a commercial trichromatic camera with two absorption filters to define image spectrally. Two images were taken for each target; so six-channel multichannel images were obtained. Three methods of spectral color reproduction were evaluated: pseudoinverse method, canonical correlation regression (CCR), and Matrix R method. The CCR method can obtain the highest spectral accuracy among these methods, just because it incorporates fifteen cross product terms in the simulation. The Matrix R method can reach the same spectral accuracy as the pseudoinverse method, and the spectral accuracy of both methods could be improved if they also use the same cross product terms. On the other hand, the Matrix R can achieve the best colorimetric accuracy for a certain combination of illuminant and observer. Thus, the Matrix R is a very promising method for achieving artwork images with sufficient spectral and colorimetric accuracy.
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Art-SI.org (Art Spectral Imaging)
RIT – Main Campus