Co-host: The Institute of Image Electronics Engineers of Japan, Cybermedia Center, Osaka University
The study on the recovery of spectral reflectances through the use of an image data has been done actively for the reproduction of a color image under a variety illuminations by using color appearance models. Since spectral reflectances are unique to objects which are independent on an illumination and on an image acquisition device, the accurate recovery of spectral reflectances is very important. The authors have already shown that the estimation of noise variance in an image acquisition system by the use of spectral reflectances of learning samples is very useful to recover spectral reflectances of test samples accurately from the test sample's image data by the use of the Weiner estimation with the estimated noise variance. In this paper, the spectral reflectances of test samples were recovered when different illuminations were used for image acquisition of learning and test samples. It is shown that recovery performance of our model is more accurate than the regression model, which is usually used, and theoretical discussion is given for this result.
