AbstractWe propose an approach for estimating and reconstructing the material appearance of objects based on the spectral image data acquired in complex illumination environments with multiple light sources. The object appearance can be constructed with various material properties such as spectral reflectance, glossiness, and matteness under different geometric and spectral illumination conditions. The objects are assumed to be made of an inhomogeneous dielectric material with gloss or specularity. The color signals from the object surface are described by the standard dichromatic reflection model, which consists of diffuse and specular reflections, where the specular component has the same spectral composition as the illuminant. The overall appearance of objects is determined by a combination of chromatic factors, based on the reflectance and illuminant spectra, and shading terms, which represent the surface geometries of the surface illumination. Therefore, the appearance of a novel object can be reconstructed by modifying the chromatic factors and shading terms. The method for appearance estimation and reconstruction comprises four steps: (1) illuminant estimation, (2) spectral reflectance estimation, (3) shading term estimation and region segmentation, and (4) appearance reconstruction based on the reflection model. The proposed approach is validated in an experiment in which objects of different materials are illuminated using different light sources. We demonstrate typical reconstruction results with novel object appearances.
Read full abstract