Spectral band Selection Using a Genetic Algorithm Based Wiener Filter Estimation Method for Reconstruction of Munsell Spectral Data

• Published in: Electronic Imaging Vol. 29; no. 18; pp. 190 - 193
• Main Authors: Ansari, Keivan, Thomas, Jean-Baptiste, Gouton, Pierre
• Format: Journal Article
• Language: English
• Published: Society for Imaging Science and Technology 29.01.2017
• Subjects: CMOS SENSOR; Gaussian Filters; Genetic Algorithm; Spectral Data Reconstruction; Wiener Estimation
• ISSN: 2470-1173; 2470-1173
• DOI: 10.2352/ISSN.2470-1173.2017.18.COLOR-059

Spectrophotometers are the common devices for reflectance measurements. However, there are some drawbacks associated with these devices. Price, sample size and physical state are the main difficulties in applying them for reflectance measurement. Spectral estimation using a set of camera-filters is the eligibly solution for avoiding these difficulties. Meanwhile band selection of filters are needed to be optimized in order to apply in imaging systems. In the present study, the Genetic algorithm was applied for finding the best set of three to eight filters combinations with specific FWHM. The algorithm tries to minimize the color difference between reconstructed and actual spectral data, assuming a simulation of imaging system. This imaging system is composed of a CMOS sensor, illuminant and 1269 matt Munsell spectral data set as the object. All simulations were done in visible spectrum. The optimized filter selections were modeled on a CMOS sensor in order to spectral reflectance reconstruction. The results showed no significant improvement after selecting a seven filter set although a descending trend in the color difference errors was obtained with increasing the number of filters.