Multispectral Analysis of Oil Paintings and Data Compression

The present paper proposes a method for estimating surface reflection properties of an oil painting from the multi-band image data and compressing the data volume in image analysis/synthesis. First we show the improved multi-band imaging system. Next we estimate the surface properties, including surface normal, surface spectral reflectance, and reflection model parameters. Then we analyze statistically the whole data set of the estimated surface-spectral reflectances. The surface spectral reflectances are represented by a linear combination of a small number of basis spectral patterns. Use of the basis patterns leads to reducing the spectral data storage. The quality of this representation is examined based on both color reproduction accuracy and memory capacity. Finally the compressed data are combined for rendering images of the paintings under arbitrary conditions of illumination and viewing