Polarization Super-Resolution Imaging Method Based on Deep Compressed Sensing

Abstract

The division of focal plane (DoFP) polarization imaging sensors, which can simultaneously acquire the target’s two-dimensional spatial information and polarization information, improves the detection resolution and recognition capability by capturing the difference in polarization characteristics between the target and the background. In this paper, we propose a novel polarization imaging method based on deep compressed sensing (DCS) by adding digital micromirror devices (DMD) to an optical system and simulating the polarization transmission model of the optical system to reconstruct high-resolution images under low sampling rate conditions. By building a simulated dataset, training a polarization super-resolution imaging network, and showing excellent reconstructions on real shooting scenes, compared to current algorithms, our model has a higher peak signal-to-noise ratio (PSNR), which validates the feasibility of our approach.