Separating reflective and fluorescent components for dynamic scenes

Separating reflective and fluorescent components by hyperspectral (HS) imaging is significant in many applications. This paper designs an imaging system, where both HS reflective images and HS fluorescent images could be obtained from the same scene, even scenes within moving objects. The system consists of a high-frequency-spectra light source and a spatially-spectrally encoded camera. During the capture phase, the light source illuminates the scene with two high-frequency lighting patterns complemented in the spectral domain by turns, then encoded camera captures a image pair accordingly. During the reconstruction phase, sparsity of the natural reflective and fluorescent HS images is utilized to recover reflective and fluorescent spectra from encoded image pair. Benefited from double-shot imaging system, dynamic scene could also be handled. The method is tested in various datasets(including synthetic and real data), and the results demonstrate that the system could achieve high-resolution hyperspectral reflectance and fluorescence recovery with high-accuracy for dynamic scenes, which can be applied for spectral relighting of real scenes.