Reconstruction of spectral reflectance based on mixed weighting and local optimization

Abstract

High-fidelity colour reproduction can be performed through reconstructing a spectral reflectance of an object surface. The particular branches in need of this reproduction range from colour printing to artistic fields. In order to improve the reconstruction accuracy for the spectral reflectance, we suggest a spectral-reflectance reconstruction method based on a mixed weighting (MW). This method is a combination of a number of earlier techniques, namely a Wiener estimation method and so-called methods of weighting on a group with smaller colour difference and weighting on a group with smaller spectral reflectance error. Specifically, we have obtained the spectral estimation with higher reconstruction accuracy by weighting the reconstruction spectra obtained by different methods. The weights for these methods have been selected through minimizing the colour difference. The MW method makes a full use of advantages of the underlying methods. It reveals high accuracy and reduces the shortcomings of those methods. Our experimental results confirm that the MW method improves the reconstruction accuracy and the stability of spectral-reflectance data.