Research on Reconstruction of Spectral Reflectance Based on Principal Component Analysis


Article Preview

Traditional color reproduction technology based on the Metamerism principle, the disadvantage is that different observer condition leads to different color appearance.To fulfill the color consistency, the spectrum reflectance of the object color sample need to be reconstructed. The principal component analysis makes use of the linear combination of a few principal components to reconstruct the spectral reflectance of sample. This paper analyzes the 31*31 matrix of Munsell spectral data by the principle component analyze method and achieves the principal component for spectrum reflectance. The numbers of principal components are identified as six by discussing the variance contribution rate. Spectral reconstruction of four Munsell testing samples makes use of first six principal components, which has met the accuracy requirements. Research shows that the reconstruction of spectral accuracy decreased when training samples and testing samples belong to the different database.



Edited by:

Ouyang Yun, Xu Min, Yang Li and Liu Xunting




Y. Zhang and S. S. Zhou, "Research on Reconstruction of Spectral Reflectance Based on Principal Component Analysis", Applied Mechanics and Materials, Vol. 262, pp. 53-58, 2013

Online since:

December 2012




[1] Giordano B. Understand color. Company Courses, Hewlett-Packard Company, (2008).

[2] Vrhel J. M. Mathematical methods of color correction. Dissertation of Ph. D, North Carolina State University, (1993).

[3] Bakke M. A., Farup I., Hardeberg Y. J. Multispectral gamut mapping and visualization— afirst attempt. Proc. SPIE-IS&T, 2005, 5667: 193-200.

[4] Day A. E. , Berns S. & , Taplin A. L. , et a1. . A psychophysical experiment evaluating the color accuracy of several multispectral image capture techniques. Journal of Imaging Science and Technology, 2004, 48(2): 93·104.

[5] Imai H. M. , Bems S. & . Spectral estimation of artist oil paints using multi—filter trichromatic imaging. Proe. 9th Congress of the International Colour Association, SPIE, 2002, 442 1: 504-507.


[6] Imai H. F. , Taplin A. L. , Day A. E. Comparison of the accuracy of various transformations from multi-band images to reflectance spectra. Technical report, Munsell Color Science Lab. (2002).

[7] Wu Chungyi, Lee Shunming, Wen Chaohua, et a1. . Multi-spectral Image acquisition system for color spectrum reproduction. Proe. 16th IPPR Conference on Computer Vision, Graphics and Image Processing, 2003, l15-122.

[8] Marimont H. D., Wandell A. B. Linear models of surface and illuminant spectra. Optical Society ofAmerica, 1992, 9(1 1): 1905-(1913).

[9] Baribeau R. . Optimized spectral estimation methods for improved colorimetry with laser scanning systems. Proc. 1st Int. Symposium on 3D Data Processing Visualization and Transmission, 2002, 29. 32.