Seung Ah Lee's Imaging Intelligence Laboratory at Seoul National University develops lensless cameras: phase masks made by grayscale lithography (K. C. Lee et al., Optica 2023) or directly on the image sensor (Y. Lee et al., IEEE Photonics J. 2022), polarization and hyperspectral lensless cameras (N. Baek et al., APL Photonics 2022; T. Kim et al., APL Photonics 2023) and physics-guided reconstruction (D. Bae et al., ECCV 2026).
A conventional camera needs a stack of lenses, which sets a lower limit on its size, weight and cost. A lensless camera replaces the lenses with a thin patterned mask just above a bare image sensor. The mask encodes the scene into a pattern on the sensor, and the image is reconstructed computationally. We design the camera hardware for each application and develop the reconstruction algorithms: masks with designed point-spread functions are made in a single lithography step, also directly on the sensor, and physics-guided neural networks reconstruct the images. Applications include polarization, hyperspectral and eye-tracking cameras, and privacy-preserving imaging.
9 projects
In a lensless camera, the mask's surface relief sets the point-spread function (PSF), the pattern that a single point of light forms on the sensor. Early lensless cameras relied on off-the-shelf diffusers or on masks that were slow and costly to make.
A phase mask must sit at a precise distance above the sensor; a small spacing error blurs the PSF and degrades the reconstruction. Instead of aligning a separate mask, we replicate it directly on the sensor. A PDMS mold taken from a master phase mask is pressed into a layer of photocurable resin on the image sensor, and the relief is fixed with UV light.
Polarization reveals stress in plastics, birefringence in biological tissue and surface properties that ordinary color images miss. Most commercial polarization cameras pattern tiny linear polarizers over the pixels, so they cannot see circular polarization, and full-Stokes instruments are usually bulky.
Hyperspectral imaging records a full spectrum at every pixel, which helps identify materials and tissue types, but most systems scan over time or need bulky dispersive optics. We placed a linear variable filter, whose passband shifts with position, at the aperture of a phase-mask lensless camera.
Classical lensless reconstruction solves an optimization problem for every frame, which is slow and leaves characteristic artifacts. In one of the lab's first lensless papers, we trained an encoder–decoder (U-Net) network to map raw measurements from our phase-mask camera directly to images.
Lensless measurements are heavily multiplexed: every pixel mixes light from across the scene, so reconstruction is ill-conditioned and sensitive to calibration errors and model mismatch. Hybrid networks that apply a physics-based deconvolution once and then hand the result to a generic neural network tend to lose track of the measurement deeper in the network.
Designing optics by gradient descent needs a wave-optics simulator that is accurate, differentiable and fast. The standard angular spectrum method (ASM) needs dense lateral grids and many thin axial slices to model thick refractive elements and steep incidence angles, which quickly becomes too expensive for wide-angle, large-aperture designs.
Microsaccades, tiny involuntary eye movements that occur even while we fixate, may carry early signs of Alzheimer's and Parkinson's disease, but measuring them usually takes high-speed, lab-grade eye trackers. Our goggle-type tracker pairs a phase mask with an inexpensive rolling-shutter CMOS sensor and infrared illumination.
A lensless camera does not have to produce pictures for people. Two directions we presented at the Optica Imaging Congress (COSI 2025) treat the mask as a design variable for what a camera should, and should not, reveal. Privacy by design: an ordinary lensless camera has a shift-invariant PSF, so its raw data can be inverted by anyone with enough example pairs.
16 publications
Integrated Forward–Inverse Network for Lensless Image Reconstruction
Bae, D., Jung, J., Kang, Y.G., Lee, K.C., Kim, T., Kim, J., Byun, S., Park, J., & Lee, S.A.
Computer Vision – ECCV 2026 (Lecture Notes in Computer Science), pp. 628–645 (2026) doi:10.1007/978-3-032-37517-9_35
Geometry-aware phase compensation for sampling-efficient angular spectrum method
Kang, Y.G., Bae, D., Jung, J., Byun, S., Ahn, H.J., Chang, H., Park, J., Lee, H., & Lee, S.A.
Optics Express, 34(8), 15244 (2026) doi:10.1364/OE.586512
Lee, H., Mun, D.J., Lee, H., Lee, K.C., Hong, S., Kim, G., Lee, S.A., & Chun, S.Y.
Image and Vision Computing, 165, 105786 (2026) doi:10.1016/j.imavis.2025.105786
Aperture-encoded snapshot hyperspectral imaging with a lensless camera
Kim, T., Lee, K.C., Baek, N., Chae, H., & Lee, S.A.
APL Photonics, 8(6), 066109 (2023) doi:10.1063/5.0150797
Design and single-shot fabrication of lensless cameras with arbitrary point spread functions
Lee, K.C., Bae, J., Baek, N., Jung, J., Park, W., & Lee, S.A.
Optica, 10(1), 72–80 (2023) doi:10.1364/OPTICA.466072
Lensless polarization camera for single-shot full-Stokes imaging
Baek, N., Lee, Y., Kim, T., Jung, J., & Lee, S.A.
APL Photonics, 7(11), 116107 (2022) doi:10.1063/5.0120465
Fabrication of Integrated Lensless Cameras via UV-Imprint Lithography
Lee, Y., Chae, H., Lee, K.C., Baek, N., Kim, T., Jung, J., & Lee, S.A.
IEEE Photonics Journal, 14(2), 1–8 (2022) doi:10.1109/JPHOT.2022.3157373
Noniterative sub-pixel shifting super-resolution lensless digital holography
Lee, H., Kim, J., Kim, J., Jeon, P., Lee, S.A., & Kim, D.
Optics Express, 29(19), 29996–30006 (2021) doi:10.1364/OE.433719
Lensless Imaging with an End-to-End Deep Neural Network
Bae, D., Jung, J., Baek, N., & Lee, S.A.
2020 IEEE International Conference on Consumer Electronics – Asia (ICCE-Asia), pp. 1–5 (2020) doi:10.1109/icce-asia49877.2020.9276865
A smartphone-based chip-scale microscope using ambient illumination
Lee, S.A., & Yang, C.
Lab on a Chip, 14(16), 3056–3063 (2014) doi:10.1039/C4LC00523F Before IIL
Imaging and identification of waterborne parasites using a chip-scale microscope
Lee, S.A., Erath, J., Zheng, G., Ou, X., Willems, P., Eichinger, D., Rodriguez, A., & Yang, C.
PLOS ONE, 9(2), e89712 (2014) doi:10.1371/journal.pone.0089712 Before IIL
Lee, S.A., Ou, X., Lee, J.E., & Yang, C.
Optics Letters, 38(11), 1817–1819 (2013) doi:10.1364/OL.38.001817 Before IIL
On-chip continuous monitoring of motile microorganisms on an ePetri platform
Lee, S.A., Zheng, G., Mukherjee, N., & Yang, C.
Lab on a Chip, 12(13), 2385–2390 (2012) doi:10.1039/C2LC40090A Before IIL
Lee, S.A., Leitao, R., Zheng, G., Yang, S., Rodriguez, A., & Yang, C.
PLOS ONE, 6(10), e26127 (2011) doi:10.1371/journal.pone.0026127 Before IIL
Zheng, G., Lee, S.A., Antebi, Y., Elowitz, M.B., & Yang, C.
Proceedings of the National Academy of Sciences, 108(41), 16889–16894 (2011) doi:10.1073/pnas.1110681108 Before IIL
Sub-pixel resolving optofluidic microscope for on-chip cell imaging
Zheng, G.*, Lee, S.A.*, Yang, S.*, & Yang, C.
Lab on a Chip, 10(22), 3125–3129 (2010) doi:10.1039/C0LC00213E Before IIL