Mid-wave infrared (MWIR) focal-plane-array (FPA) compressive imaging uses a digital micromirror device (DMD) and computational reconstruction to give a small, inexpensive infrared detector a large effective pixel count. The state-of-the-art system of Jin et al. (Optics and Laser Technology 2023) achieves 4×4 pixel-level super-resolution on long-distance targets simulated by a collimator, but its 2×2 DMD super-pixel is deliberately matched to the ~17 µm diffraction-limited spot of the relay optics, so the achieved resolution reaches — and stops at — the diffraction limit of the system, and operation has only been demonstrated under laboratory conditions. The authors state three open tasks: demonstrating the system on actual long-distance (outdoor) targets where atmospheric propagation, background clutter, and temporal instability demand robust, updatable calibration; accelerating reconstruction, whose cost scales with pixel count; and, most fundamentally, achieving super-resolution beyond the diffraction limit of the relay optics. Deep-learning methods have substantially advanced reconstruction speed and calibration efficiency, and online-calibration concepts exist, but no work demonstrates MWIR FPA compressive super-resolution on real outdoor long-distance scenes, and no work in the DMD-based FPA compressive architecture resolves detail beyond its optical diffraction limit. The problem — pushing this system class from diffraction-limited laboratory imaging to deployable, beyond-diffraction-limit long-range MWIR imaging — remains open.

Background

Mid-wave infrared (MWIR) imaging covers the 3–5 µm atmospheric transmission band and underpins long-range surveillance, remote sensing, and industrial monitoring because it penetrates haze and works day and night. Large-format MWIR focal plane arrays (FPAs — the 2D detector chips that convert infrared radiation into pixelated images) are extremely expensive, so in most MWIR systems the small number of detector pixels, not the optics, is the binding resolution limit. Focal-plane-array-based compressive imaging (FPA-CI), developed from compressed sensing (the theory that sparse signals can be recovered from far fewer linear measurements than Nyquist sampling requires) and the single-pixel camera, removes this limit: the scene is imaged onto a digital micromirror device (DMD — a chip of individually switchable micromirrors, each 10.8 µm across in the system considered here), which multiplies the scene by a sequence of high-resolution binary masks (e.g., Hadamard patterns, a standard family of orthogonal two-level black/white patterns); a low-resolution detector records one multiplexed frame per mask, and the high-resolution image is recovered computationally by solving an inverse problem.

The mapping between DMD mirrors and detector pixels — the calibration or transfer-function matrix — must be measured experimentally; point-by-point scanning calibration is slow, low-flux, and sensitive to thermal drift, which motivated Jin et al.'s compressive calibration, in which the calibration matrix itself is recovered from few high-flux random binary-mask measurements. Jin et al. (Optics and Laser Technology 157, 108740, 2023) built a long-distance MWIR FPA-CI system: a blackbody source and objects placed at the view port of a 750 mm telescope produce collimated light that simulates targets at infinity; a DMD modulates the scene and relay optics demagnify it onto a 68×68-pixel region of an MWIR camera, from which 256×256-pixel images are reconstructed. Because the relay optics' diffraction-limited resolution (~17 µm, the smallest spot to which the optics can focus light, set by the finite aperture and the 3–5 µm wavelength) is comparable to the DMD mirror pitch (10.8 µm), a 2×2-mirror group ('super-pixel', about 21.6 µm across) is the finest modulation unit matched to the diffraction spot, and the realized 4×4 resolution enhancement exactly exhausts this modulation level — the system's resolution ceiling is the diffraction limit, not the detector.

Follow-up work by the same and a sibling group has pushed reconstruction speed (GPU acceleration reaching 20 frames per second in a 2021 Optics Letters MWIR camera; a physics-enhanced deep network in 2023–2024 achieving real-time 4×4 super-resolution with three designed masks) and calibration quality (a 2025 high-order physical model that includes diffraction, aberration, and misalignment as degrading factors with one-shot learned calibration). A 2025 invited review of pixel super-resolution imaging via spatio-temporal encoding modulation (Acta Optica Sinica 45, 1720013) confirms the field-level picture: pixel super-resolution defeats the sampling limit, but when the diffraction spot exceeds the pixel size, no oversampling recovers frequencies beyond the optical cutoff, and robust operation under atmospheric turbulence and thermal drift — requiring dynamic online calibration — remains an open systems challenge.

Problem Statement

For DMD-based focal-plane-array compressive imaging in the mid-wave infrared with compressive calibration, as established by Jin et al. (Optics and Laser Technology 157, 108740, 2023), extend the demonstrated capability from laboratory conditions — collimator-simulated targets at infinity, 4×4 pixel-level super-resolution that exactly reaches the diffraction-limited modulation level set by the 2×2 DMD super-pixel and the ~17 µm diffraction spot of the relay optics — to a deployable long-distance imaging system. The stated open objectives are: (1) demonstrate super-resolution imaging of actual long-distance objects, confronting the stronger and time-varying environmental noise of real atmospheric propagation (turbulence, absorption, background clutter, temporal instability), which requires robust calibration and calibration-update strategies under field conditions; (2) improve the efficiency of the computational reconstruction, whose speed is restricted by the imaging pixel number, so that the system scales to high pixel counts; and (3) — the fundamental ceiling — achieve super-resolution beyond the diffraction limit of the relay optics, i.e., resolve object detail finer than the optical point-spread limit that the current super-pixel design is matched to, by whatever encoding, acquisition, or reconstruction principle makes spatial-frequency content beyond the optical cutoff recoverable in this architecture. These objectives jointly serve one overarching goal: a low-cost MWIR imaging capability whose resolution is no longer capped either by detector pixel count or by the diffraction limit of its optics, operating on real long-distance scenes.

References

  1. Xiao-Peng Jin, An-Dong Xiong, Xiao-Qing Wang, Xu-Ri Yao, Xue-Feng Liu, Qing Zhao, 'Long-distance mid-wave infrared super-resolution compressive imaging', Optics and Laser Technology 157 (2023) 108740, DOI: 10.1016/j.optlastec.2022.108740, https://doi.org/10.1016/j.optlastec.2022.108740 (arXiv preprint: Xiao-Peng Jin, Qing Zhao, Xue-Feng Liu, An-Dong Xiong, 'Mid-wave infrared super-resolution imaging based on compressive calibration and sampling', arXiv:2109.04887, https://arxiv.org/abs/2109.04887)
  2. Linxia Zhang, Jun Ke, Sheng Chi, Xiaowen Hao, Tong Yang, Dewen Cheng, 'High-resolution fast mid-wave infrared compressive imaging', Optics Letters 46 (2021) 2469–2472, DOI: 10.1364/ol.420481, https://doi.org/10.1364/ol.420481
  3. Xiao-Peng Jin, An-Dong Xiong, Fan Liu, Xiao-Qing Wang, Wei Zhang, Chang-Heng Li, Xu-Ri Yao, Xue-Feng Liu, Qing Zhao, 'Parallel compressive super-resolution imaging with wide field-of-view based on physics enhanced network', Applied Physics Letters 124 (2024) 211106, DOI: 10.1063/5.0202195, https://doi.org/10.1063/5.0202195 (arXiv:2310.14965, https://arxiv.org/abs/2310.14965)
  4. Jingwen Lei, Xu Ma, Xianhong Zhao, Zhen Fang, Xiaowen Hao, Junyao Zhao, Jun Ke, 'Super-resolution infrared compressive imaging based on a high-order physical model', Chinese Optics Letters 23 (2025) 111105, DOI: 10.3788/col202523.111105, https://doi.org/10.3788/col202523.111105
  5. Kunyao Liang, Xu Zhang, Zihao Pei, Hongchun Li, Xin Liu, Bowen Wang, Qian Chen, Chao Zuo, 'Pixel super-resolution imaging based on spatio-temporal encoding modulation: current status and future trends (invited)' (Chinese title: 基于时空编码调控的像素超分辨成像：研究现状与发展趋势（特邀）), Acta Optica Sinica 45 (17) (2025) 1720013, DOI: 10.3788/aos251177, https://doi.org/10.3788/aos251177 (in Chinese; English title from the journal's bilingual record)
  6. Ting Wang, Chao Wang, Xinkai Wu, Hongyu Sun, Jianan Liu, 'Infrared super-resolution imaging using minimalist optical systems' (Chinese title: 极简光学系统红外超分辨率成像), Acta Optica Sinica 45 (2025) 1111002, DOI: 10.3788/aos250541, https://doi.org/10.3788/aos250541 (in Chinese; English title from the journal's bilingual record)
  7. Yinran Shen, Xuri Yao, Shijian Li, Chao Shen, Yuhao Wang, Chongwu Shao, Qing Zhao, 'Parallel single-pixel imaging based on modulation region expansion and overlapping reconstruction', arXiv:2608.17579 (2026), https://arxiv.org/abs/2608.17579
  8. Alper Güngör, M. Umut Bahceci, Yasin Ergen, Ahmet Sözak, O. Oner Ekiz, Tolga Yelboga, Tolga Çukur, 'CalibFPA: A focal plane array imaging system based on online deep-learning calibration', IEEE Transactions on Computational Imaging 10 (2024) 1650–1663, DOI: 10.1109/TCI.2024.3477312, https://doi.org/10.1109/TCI.2024.3477312 (arXiv:2309.11421, https://arxiv.org/abs/2309.11421)
  9. Junyao Zhao, Xiaowen Hao, Xu Ma, Jun Ke, 'Resolution enhancement via an unfolding network with a deblocking module and DCN for infrared small block-based compressive imaging', Chinese Optics Letters 24 (2026) 011103, DOI: 10.3788/col202624.011103, https://doi.org/10.3788/col202624.011103
  10. Xiaowen Hao, Linxia Zhang, Xu Ma, Jun Ke, 'Decoupled sensing matrices with swin transformer for high speed and resolution imaging', Optics and Laser Technology 204 (2026) 116001, DOI: 10.1016/j.optlastec.2026.116001, https://doi.org/10.1016/j.optlastec.2026.116001
  11. Zunwang Bo, Chenjin Deng, Fei Wang, Wenlin Gong, Yuanhao Su, Yichen Zhang, Mingliang Chen, Chunfang Wang, Shensheng Han, 'High-resolution long-range 3D single-photon imaging with a compact SPAD array', arXiv:2604.08835 (2026), https://arxiv.org/abs/2604.08835
  12. Zaihong Xiao, Zhenghuan Li, Guan Wang, Yihao Xue, Jiahao Li, Yiqi Jia, Yu Cai, Huaxia Deng, Xinglong Gong, 'Aperture sampling and coarse-to-fine reconstruction for seamless parallel compressive sensing imaging', Optics and Lasers in Engineering 201 (2026) 109716, DOI: 10.1016/j.optlaseng.2026.109716, https://doi.org/10.1016/j.optlaseng.2026.109716