Single-Pixel Multispectral Imaging With Customized Pixelated Spectral Mask Plate

Single-pixel imaging (SPI) has emerged as a cutting-edge computational technique, offering distinct advantages over conventional methods, particularly in non-visible wavelengths. This study introduces a compact and low-cost single-pixel multispectral imaging (SP-MSI) system that utilizes a customized pixelated spectral mask plate (PSMP) for efficient spectral modulation. By employing pre-calibrated multispectral patterns, our method enables the reconstruction of a 3D datacube ((Formula presented.)) at sampling ratios below 30%. This is achieved through a custom-designed deep unfolding network (DUN) for image reconstruction. We validate the performance of our system through comprehensive theoretical analysis and experimental demonstrations across eight spectral bands. The proposed approach presents a promising solution for developing efficient and compact multispectral and hyperspectral imaging systems, compatible with both single-pixel detectors and conventional 2D sensors.