Improving Training-based Reflectance Reconstruction via White-balance and Link Function

An improved spectral reflectance reconstruction method is developed to transform camera RGB to spectral reflectance by inserting white balance and link function during the training-based method. The novelty in our method is the use of white-balancing to normalize the scene illumination and link function to transform the reflectance, we use a radial basis function network to model the mapping between camera-specific RGB values and specific reflectance spectra. Experimental results indicate that the proposed method significantly outperforms currently existing methods in terms of spectral error and shape especially under the illumination not present in the training process.