Niloofar ESLAHI, Seyed Hossein AMIRSHAHI^*, and Farnaz AGAHIAN

Department of Textile Engineering, Amirkabir University of Technology, Tehran 15914, Iran

(Received December 24, 2008; Revised March 9, 2009; Accepted March 13, 2009)

The weighted canonical correlation regression technique is employed for reconstruction of reflectance spectra of surface colors from the related XYZ tristimulus values of samples. Flexible input data based on applying certain weights to reflectance and colorimetric values of Munsell color chips has been implemented for each particular sample which belongs to Munsell or GretagMacbeth Colorchecker DC color samples. In fact, the colorimetric and spectrophotometric data of Munsell chips are selected as fundamental bases and the color difference values between the target and samples in Munsell dataset are chosen as a criterion for determination of weighting factors. The performance of the suggested method is evaluated in spectral reflectance reconstruction. The results show considerable improvements in terms of root mean square error (RMS) and goodness-of-fit coefficient (GFC) between the actual and reconstructed reflectance curves as well as CIELAB color difference values under illuminants A and TL84 for CIE1964 standard observer.

Key words: canonical correlation analysis, canonical correlation regression, principal component analysis, colorimetric data, recovery of spectral reflectance

^*E-mail address: hamirsha@aut.ac.ir