High-Efficiency Multispectral-Polarization Imaging System Using Polarization Camera Array With Notch Filters

Incident light captured by an optical imaging system engenders multidimensional high-level optical information, such as spatial resolution, light intensity, spectrum, and polarization. However, panchromatic image sensors can only capture spatial and light intensity information. This article presents a novel multispectral-polarization imaging approach that overcomes the limitations related to low light flux and slow imaging speed found in multispectral-polarization techniques based on bandpass filters. By utilizing notch filters and a camera array and integrating computational imaging algorithms, this method achieves high temporal, spatial, and spectral resolution in polarization imaging for the first time. The spatial, light intensity, spectral, and polarization information of the target scene can be obtained synchronously while guaranteeing high light efficiency performance of the system. To obtain high spectral resolution polarization images, a K-times singular value decomposition (K-SVD) algorithm is used to extract the feature dictionary of the hyperspectral dimension. Then, a compressive sensing-based algorithm is applied to conduct spectral super-resolution. The principle of compressive sensing is combined with prior sparsity and smoothness information to perform spectral super-resolution. The performance of the proposed multispectral-polarization imaging system and the effectiveness of the proposed method were verified using a public multispectral dataset and field experiments. The experimental results confirm that the proposed technology can obtain high-quality spatial resolution, light intensity, spectral, and polarization data of the target scene in a single shot. The algorithm can reconstruct 16 spectral band images from four captured spectral images while adequately preserving the polarization information of the target scene in the calculated spectral imaging results.