Multispectral Constancy Based on Spectral Adaptation Transform

The spectral reflectance of an object surface provides valuable information about its characteristics. Reflectance reconstruction from multispectral images is based on certain assumptions. One of these assumptions is that the same illumination is used for system calibration and image acquisition. We propose the novel concept of multispectral constancy, achieved through a spectral adaptation transform, which transforms the sensor data acquired under an unknown illumination to a generic illuminant-independent space. The proposed concept and methods are inspired from the field of computational color constancy. Spectral reflectance is then estimated by using a generic linear calibration. Results of reflectance reconstruction using the proposed concept show that it is efficient, but highly sensitive to the accuracy of illuminant estimation.