Comparative study on three reconstruction algorithms for a single pixel hyperspectrometer

In this paper, we introduce a single-pixel hyperspectral system based on liquid crystal variable phase retarder (LCVR). Compressed sensing reconstruction algorithms L1-magic, Two-Step Iterative Shrink-age/Thresholding (TwIST), Gradient Projection for Sparse Reconstruction (GPSR) for single-pixel hyperspectral reconstruction are compared. The root mean square error (RMSE), peak signal-to-noise ratio (PSNR), correlation coefficients (CC) and convergence time (CT) are used as the evaluation criterions to estimate the performance of algorithms under various compression ratio. It helps us to choose the optimal algorithm to different application cases.