Recovering Spectral Reflectances of Artistic Paintings Through the Use of Image Data and Evaluating their Accuracies

摘要

Recovery of spectral reflectances of artistic paintings through the use of image data at the pixel resolution is very important for reproducing color images under various illuminations. Image data of three types of pictures painted by different colorants, such as water colors, oil colors and natural minerals, were captured by a multispectral camera and the image data were used for recovering spectral reflectances by the adaptable least square error method. Usually colors of paintings are not uniform within the area used to measure a spectral reflectance by a spectrophotometer and the area contains many pixels within it. It was already shown that the accuracies of the recovered spectral reflectances of a painting must be evaluated by averaging the spectral reflectances recovered by the pixels within the area used for the measurement. For the first time the new evaluation method aforementioned was used for the spectral reflectances of artistic paintings recovered by the adaptive recovery method. The performance of this method was compared with the previously proposed model and it is shown that the adaptive method is effective in improving the recovery performance.