Lune

NeurIPS2025Top-tier venue

Dual-Comb Ghost Imaging with Transformer-Based Reconstruction for Optical Fiber Endomicroscopy

David Dang, Myoung-Gyun Suh, Maodong Gao, Byoung-Jun Park, Beyonce Hu, Yucheng Jin, Wilton J. M. Kort-Kamp, Ho Wai Lee

Abstract

Endoscopic imaging is indispensable for visualizing internal organs, yet conventional systems remain bulky and costly because they rely on large, multi-element optics, which limits their ability to access and image certain areas of the body. Achieving high-quality endomicroscopy with hundred micron-scale and inexpensive hardware remains a grand challenge. Optical fibers offer a sub-millimeter-scale imaging conduit that could meet this need, but existing fiber-based approaches typically require either raster scanning or multicore bundles, which limit resolution and speed of imaging. In this work, we overcome these limitations by combining dual-comb interferometry with optical ghost imaging and advanced algorithm. Optical frequency combs enable precise and parallel speckle illumination via wavelength-division multiplexing through a single-core fiber, while our dual-comb compressive ghost imaging approach enables snapshot detection of bucket-sum signals using a single-pixel detector, eliminating the need for both spatial and spectral scanning. To reconstruct images from these highly compressed measurements, we introduce Optical Ghost-GPT , a transformer-based image reconstruction model that enables fast, high-fidelity recovery at low sampling ratios. Our dual-comb ghost imaging approach, combined with the novel algorithm, outperforms classical ghost imaging techniques in both speed and accuracy, enabling real-time, high-resolution endoscopic imaging with a significantly reduced device footprint. This advancement paves the way for non-invasive, high-resolution, low-cost endomicroscopy and other sensing applications constrained by hardware size and complexity.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

Questions to start from

Lune can draw on

the full paper · related work

Get the answer

Free to start. No credit card required.

Lineage

Builds onCited byLeft to right by year; the more cited, the higher.

Builds on1

  1. Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image EncodingPengchuan Zhang, Xiyang Dai, Jianwei Yang, Bin Xiao et al.ICCV 2021 · 384 citations

Related papers

  1. Spectrum from Defocus: Fast Spectral Imaging with Chromatic Focal StackM. Kerem Aydin, Yi-Chun Hung, Jaclyn Pytlarz, Qi Guo et al.CVPR 2026 · 1 citation
  2. DepthScopy: Decoupling Frequency for Endoscopic Depth Estimation in Sparsely-Textured RegionsMinghai Shi, Xiaoxian Zhang, Xiaoyue Liu, Fan Yang, Lei LiCVPR Findings 2026
  3. FourierNets enable the design of highly non-local optical encoders for computational imagingDiptodip Deb, Zhenfei Jiao, Ruth R. Sims, Alex B. Chen et al.NeurIPS 2022 · 20 citations
  4. DetectSCI: Toward Object-Guided ROI Reconstruction for High-Resolution Video Snapshot Compressive ImagingXingjian Jiang, Lishun Wang, Ping Wang, Xin YuanCVPR 2026
  5. Metascope: Optics-Driven Neural Network for Ultra-Micro Metalens EndoscopyWuyang Li, Wentao Pan, Xiaoyuan Liu, Zhendong Luo et al.ICCV 2025 · 1 citation
Dusk over the sea between two cliffs drawn in fine vertical lines