Hybrid Single-Pixel Camera for Dynamic Hyperspectral Imaging

Single-pixel imaging recovers an image from a sequence of point measurements that correspond to inner products between the image of the scene and some light patterns. This technique is particularly suited to hyperspectral imaging which has a wide range of biomedical applications, in particular fluorescence-guided neurosurgery where high spectral resolution is required to improve tumor resection. However, the sequential nature of the acquisition leads to strong artifacts when a dynamic scene is considered. This paper proposes a simple yet efficient strategy for imaging such dynamic scenes. Our method relies on a hybrid device that combines a standard imaging arm with a hyperspectral single-pixel imaging arm. By estimating the motion from the standard imaging arm, we show how to compensate the motion perceived by the hyperspectral arm. We demonstrate the efficiency of the proposed strategy on simulated and experimental data.