Effect of the restoration of saturated signals in hyperspectral image analysis and color reproduction

In hyperspectral imaging, the captured signal is often affected by saturation due to specular reflection or a peaky spectrum. In this paper, we propose a restoration method for saturated hyperspectral signals. Our algorithm is based on principal component analysis to obtain the reconstruction basis and then solve a linear constrained least square problem to calculate the coefficients of each basis. We discuss the problems that saturated signals might cause and apply our method to two sets of real hyperspectral images and a set of hyperspectral images with simulated saturation. The results show that our method helps increase unsupervised object detection and improves high-fidelity color reproduction.