Top

Spatially Varying Point Spread Function Conditioned Swin Transformer for Lensless Image Reconstruction

Authors

Files

PDF

Abstract

Lensless cameras encode scene radiance via a thin optical element; due to the spatial variation of the point spread function, a single global blur model is insufficient for accurate reconstruction over the whole field of view. This study presents a reconstruction framework for static, approximately planar scenes that conditions a hierarchical Swin Transformer on a calibrated spatial response field. A compact optical encoder generates feature modulation signals from locally interpolated response descriptors and shifted-window attention facilitates the exchange of image information among adjacent areas. Measurement-consistency correction restricts the reconstructed image by employing the same spatially-varying forward operator for calibration. A reproducible evaluation design separates scene content, calibration positions and perturbation levels for the examination of fidelity, peripheral behavior and sensitivity to optical mismatch. The reported comparison gives the conditioned model a peak signal-to-noise ratio of 29.84 dB and structural similarity of 0.914, compared with 27.63 dB and 0.872 for the matched unconditioned Swin model. Peripheral reconstruction error decreases from 0.091 to 0.059, while the medium configuration uses 11.8 million parameters and has a latency of 43 ms. These metrics characterize image fidelity, peripheral accuracy, and computational demand within a common evaluation framework. The framework explicitly incorporates spatial calibration into the reconstruction input and enables separate evaluation of optical adaptation and image prior learning. Experiments on independently calibrated hardware are required to establish practical accuracy and transferability.

PDF

References

Zhao, Y., Xu, F., Liu, J., & Guo, C. (2025). Portable lensfree single-frame imaging sensor using random phase modulation. Optics Express, 33(26), 55054–55067.https://doi.org/10.1364/oe.582880

Li, S., Gao, Y., Wu, J., Wang, M., Huang, Z., Chen, S., & Cao, L. (2025). Lensless camera: Unraveling the breakthroughs and prospects. Fundamental Research, 5(4), 1725–1736.https://doi.org/10.1016/j.fmre.2024.03.019

Lin, Y., Li, G., Yan, D., Zhang, W., Zhu, J., & Zhao, Y. (2025). Lensless diffraction imaging using a rotated slit modulation. Optics Continuum, 4(2), 421–437.https://doi.org/10.1364/optcon.554179

Chien, T., Cao, R., Liu, F. L., Kabuli, L. A., & Waller, L. (2024). Space-time reconstruction for lensless imaging using implicit neural representations. Optics Express, 32(20), 35725–35732.https://doi.org/10.1364/oe.530480

Feshki, M., Godin, A. G., De Koninck, Y., & Gosselin, B. (2026). Learned spatially varying microscopy model with adaptive point spread functions. Optics Express, 34(7), 12287–12303.https://doi.org/10.1364/oe.586982

Chen, K., Li, Y., Li, Z., Hu, J., Zhao, Q., & Guo, Y. (2026). High-quality reconstruction method for miniaturized lensless cameras based on imaging model optimization. Optics Express, 34(15), 26909–26925.https://doi.org/10.1364/oe.595581

Qian, H., Ling, H., & Lu, X. (2024). Robust unrolled network for lensless imaging with enhanced resistance to model mismatch and noise. Optics Express, 32(17), 30267–30283.https://doi.org/10.1364/oe.531694

Liang, J., Cao, J., Fan, Y., Zhang, K., Ranjan, R., Li, Y., Timofte, R., & Van Gool, L. (2024). VRT: A video restoration transformer. IEEE Transactions on Image Processing, 33, 2171–2182.https://doi.org/10.1109/tip.2024.3372454

Kang, X., Duan, P., Li, J., & Li, S. (2024). Efficient Swin Transformer for remote sensing image super-resolution. IEEE Transactions on Image Processing, 33, 6367–6379.https://doi.org/10.1109/tip.2024.3489228

Jiang, R., & Han, Y. (2024). Generalizing to out-of-sample degradations via model reprogramming. IEEE Transactions on Image Processing, 33, 2783–2794.https://doi.org/10.1109/tip.2024.3378181

Wang, Z., Zheng, S., Ding, Z., & Guo, C. (2024). Dual-constrained physics-enhanced untrained neural network for lensless imaging. Journal of the Optical Society of America A, 41(2), 165–173.https://doi.org/10.1364/josaa.510147

Nakamura, T., Kato, R., Iwata, K., Makihara, Y., & Yagi, Y. (2024). Multilayer lensless camera for improving the condition number. Applied Optics, 63(28), G9–G17.https://doi.org/10.1364/ao.521126

Banerjee, A., & Singh, S. (2026). Towards physics-informed cyclic adversarial multi-PSF lensless imaging. IEEE Transactions on Emerging Topics in Computational Intelligence, 10(1), 390–400.https://doi.org/10.1109/tetci.2025.3579385

Ma, Z., Wang, Y., Tohidypour, H. R., Nasiopoulos, P., & Leung, V. C. M. (2024). Enhancing image quality by reducing compression artifacts using dynamic window Swin Transformer. IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 14(2), 275–285.https://doi.org/10.1109/jetcas.2024.3392868

Cui, Y., Ren, W., Cao, X., & Knoll, A. (2024). Revitalizing convolutional network for image restoration. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(12), 9423–9438.https://doi.org/10.1109/tpami.2024.3419007

Cui, Y., Zhu, J., & Knoll, A. (2025). Enhancing perception for autonomous vehicles: A multi-scale feature modulation network for image restoration. IEEE Transactions on Intelligent Transportation Systems, 26(4), 4621–4632.https://doi.org/10.1109/tits.2025.3538485

Liu, Z., Zeng, T., Zhan, X., Zhang, X., & Lam, E. Y. (2025). Generative approach for lensless imaging in low-light conditions. Optics Express, 33(2), 3021–3039.https://doi.org/10.1364/oe.544875

Ekmekci, C., & Cetin, M. (2026). Conformalized generative Bayesian imaging: An uncertainty quantification framework for computational imaging. IEEE Transactions on Computational Imaging, 12, 216–229.https://doi.org/10.1109/tci.2025.3649389

Ekec, K., & Teğin, U. (2026). Resolution-agnostic lensless imaging via Fourier neural operators. Optics Express, 34(18), 34757–34772.https://doi.org/10.1364/oe.610853

Mo, C., Liu, X., Tong, J., Xi, J., Yu, Y., & Cai, Z. (2024). Lensless light-field imaging using LMI. Optics Express, 32(22), 38112–38127.https://doi.org/10.1364/oe.539021

Westphal, M., & Zapf, A. (2024). Statistical inference for diagnostic test accuracy studies with multiple comparisons. Statistical Methods in Medical Research, 33(4), 669–680.https://doi.org/10.1177/09622802241236933

Lan, J., Xu, C., Pang, H., Yang, C., Zhao, L., Hu, S., & Yang, Y. (2025). Robust lensless computational imaging via nonlinear weighting approach. Applied Optics, 64(2), 423–430.https://doi.org/10.1364/ao.543766

Li, C., Cao, J., Hao, Q., Song, C., Yao, H., & Wei, Z. (2026). Reference-free per-measurement assessment of lensless reconstruction quality under imperfect PSF models. Optics Express, 34(23), 44252–44267.https://doi.org/10.1364/oe.613012

Kabuli, L. A., Pinkard, H., Markley, E., Hung, C. S., & Waller, L. (2026). Designing lensless imaging systems to maximize information capture. Optica, 13(2), 227–235.https://doi.org/10.1364/optica.570334

Ni, C., Yang, C., Zhang, X., Li, Y., Zhang, W., Zhai, Y., He, W., & Chen, Q. (2024). Address model mismatch and defocus in FZA lensless imaging via model-driven CycleGAN. Optics Letters, 49(15), 4170–4173.https://doi.org/10.1364/ol.528502

Fujiwara, S., Kawachi, H., Silva Neto, J. R. C. S. A. V., Yagi, Y., & Nakamura, T. (2026). Deep lensless imaging with non-iterative output-constraint satisfaction. Optics Express, 34(20), 37137–37151.https://doi.org/10.1364/oe.608276

Goswami, S., Krishnan, G., & Javidi, B. (2024). Robustness of single random phase encoding lensless imaging with camera noise. Optics Express, 32(4), 4916–4930.https://doi.org/10.1364/oe.510950

Li, Y., Li, Z., Chen, K., Guo, Y., & Rao, C. (2024). 3D reconstruction from focus for lensless imaging. Applied Optics, 63(31), 8212–8220.https://doi.org/10.1364/ao.540257

Grant-Jacob, J. A., & Mills, B. (2026). Enhancing lensless imaging through Fisher-information-engineered structured scattering. Optics Continuum, 5(7), 2249–2258.https://doi.org/10.1364/optcon.599341

Sugimura, D., Maeda, Y., Yamamura, K., Kurihara, K., & Hamamoto, T. (2025). Lensless facial image identification through joint learning of reconstruction and recognition. Optics Express, 33(14), 30384–30400.https://doi.org/10.1364/oe.565829

Bezzam, E., Perron, Y., & Vetterli, M. (2025). Towards robust and generalizable lensless imaging with modular learned reconstruction. IEEE Transactions on Computational Imaging, 11, 213–227.https://doi.org/10.1109/tci.2025.3539448

Details