Dual-Domain Shape–Polarization Optical Encoding for Detector-Multiplexed Wide-Field Infrared Small-Target Imaging

Abstract

Detector-multiplexed wide-field infrared imaging reuses detector area by folding several sub-fields of view onto a common focal-plane region, but the branch identity of an aliased small target is then unavailable from intensity alone. We propose a dual-domain optical code that combines a fixed Zemax-derived point-spread-function (PSF) library with a conditionally assigned linear-polarization codebook. To replace the earlier feature-level validation, we developed an image-level radiometric simulator that generates four analyzer images under a fixed incident-photon budget, applies mirror transfer, throughput loss, Poisson shot noise, background, read noise, and physical multi-target PSF superposition, and then re-extracts morphology and Stokes observables. The 1037 PSF samples were divided by sub-field into construction, calibration, and locked test sets of 623, 207, and 207 samples. The polarization-to-field assignment was selected by an exhaustive 9! search on the construction data, whereas the compound weight and separability threshold were fixed from calibration data only. In 25,000 locked target-level tests, the equal-weight-form compound decoder achieved 95.936% accuracy (95% confidence interval of 95.684–96.174%), compared with 10.836%, 45.564%, and 88.720% for intensity-only, shape-only, and polarization-only decoding. At the calibrated threshold, the effective capacities were five, four, and nine for the shape, polarization, and compound code spaces. The results are simulation-phase evidence and require experimental verification with measured infrared throughput and component tolerances.