An LED-Based Spectral Imaging System for Surface Reflectance and Normal Estimation

An LED-based spectral imaging system is proposed for estimating surface spectral reflectances and surface normals of art painting. The imaging system is decomposed into a multi-spectral lighting system and a digital camera. The previous lighting system has some essential limitation and inconvenience, which includes low luminance, time consumption, and expensive cost. For the purpose of solving the problems, this paper presents an LED-based spectral lighting system. This is an active illumination system in which spectral light can be controlled by a computer. We note that LED lamps with different peak wavelengths are available recently and the band width of LED light is narrow. The luminance level of emitted light is much higher and the image acquisition is much faster by changing spectral light quickly. First, we construct a flat lighting system using six LEDs with different peak wavelengths to estimate surface spectral reflectances. The surface spectral reflectances of objects are recovered from noisy spectral camera data by the Wiener estimator. Next, we construct a dome lighting system using several sets consisting of seven LEDs, mainly to estimate surface normals of art paintings. The photometric stereo technique is used for estimating the surface normal vector at every pixel point. The feasibility of the proposed method is examined in experiments using a Mac Beth Color Checker chart and an oil painting.