Optics-computation co-design for simplified DMD-based infrared super-resolution imaging

We present a system-oriented design-and-reconstruction strategy for simplified digital micromirror device (DMD)-based infrared super-resolution imaging. To reduce optical complexity, we simplify the projection optics through a theory-guided design strategy and characterize the resulting field-dependent degradation using point spread functions (PSFs) estimated at representative field points to model spatially varying blur. A block-wise approximation is adopted to make the spatially variant restoration computationally tractable. Building on this model, we develop T-L (TVAL3-Lucy-Richardson), a block-based two-stage reconstruction procedure in which TV-regularized reconstruction is first used to recover high-resolution content from coded measurements under the 2× super-resolution setting, followed by Lucy-Richardson deconvolution for residual blur correction. Relative to images acquired directly with the simplified optics, simulations show average improvements of 78% in peak signal-to-noise ratio (PSNR) and 71% in structural similarity index measure (SSIM), and laboratory experiments yield corresponding gains of 58.5% and 57.1% with clearer resolution-bar separability. Outdoor tests on representative long-range scenes further demonstrate the practical potential of the proposed method for recovering fine details. In addition to improving image quality, the simplified optical design reduces element count and cost while improving throughput and compactness, indicating the practical potential of this strategy for compact infrared platforms under size and cost constraints.