Yuyuan Han 韩聿源

Computational imaging & robust image-free sensing

Yuyuan Han

PhD student

Xi'an Jiaotong University

Academic CVPDF Internship RésuméPDF

hanyuyuan6@gmail.com

Teaching machines to see from raw measurements, not images.

I work on single-pixel imaging and image-free sensing: I design optical measurements and train networks that operate directly on the measurement sequence, without first reconstructing an image. A single-pixel camera illuminates a scene with structured patterns and records one value per pattern with a single detector; I study how much a machine can perceive from those values alone. The results hold where conventional approaches struggle: sparse Hadamard patterns keep imaging robust through scattering media, and image-free segmentation maintains near-constant accuracy down to 0.78% sampling. At Politecnico di Milano, I plan to work on time-domain diffuse optical tomography: 3D reconstruction inside scattering media from photon arrival times, from simulation to phantoms. My long-term aim is to enable robotic systems to infer physical parameters and interpret their surroundings directly from raw measurements, guided by physical models.

I am open to research collaborations and internships in computational imaging, image-free sensing and machine perception.

Outside the lab: basketball, fitness and travel.

News

Selected research

All publications
Complete STSF+TPLS architecture showing acquisition and temporal encoding, content-adaptive lift, spatial decoding, loss composition, training schedule and auxiliary-head comparisons.
Fig. 1 · Full size

The Lift Spectrum: How Measurement-to-Space Adaptivity Shapes Robustness in Image-Free Single-Pixel Sensing

Yuyuan Han, Jingwei Li, Xiaoxia Zhang, Long Qiu, Chong Wang, Wenxuan Hao, Jiangyu Han, Xinyu Yao, Yuchen He, Hui Chen, Jianbin Liu, Huaibin Zheng

arXiv preprint · 2026

Image-free sensing moves the hard problem to the 1D-to-2D lift; how adaptive that lift is determines robustness.

PaperCode
BibTeX
@misc{han2026lift,
  title = {{The Lift Spectrum: How Measurement-to-Space Adaptivity Shapes Robustness in Image-Free Single-Pixel Sensing}},
  author = {Han, Yuyuan and Li, Jingwei and Zhang, Xiaoxia and Qiu, Long and Wang, Chong and Hao, Wenxuan and Han, Jiangyu and Yao, Xinyu and He, Yuchen and Chen, Hui and Liu, Jianbin and Zheng, Huaibin},
  year = {2026},
  eprint = {2607.22077},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2607.22077}
}
  • Near-constant accuracy down to 0.78% sampling
  • Up to +21.7 mIoU at 20 dB over clean-trained reconstruct-then-segment
  • +3.2–9.9 mIoU over the prior image-free method (SPIFS)
  • Runs on a real DMD bench without fine-tuning
Complete SHSI experiment figure with two ground-glass occluders, the testing-board target and six reconstruction rows at all eight sampling rates, retaining every original panel label and comparison border.
Fig. 9 · Full size

Generalized sparse Hadamard single-pixel Imaging

Yuyuan Han, Peng Wu, Hanxiong Xu, Bin Li, Yuchen He, Jianbin Liu, Hui Chen, Huaibin Zheng

Optics Express · 2025

Orthogonal Hadamard patterns with tunable sparsity make single-pixel imaging robust through scattering and noise.

PaperCode
BibTeX
@article{han2025generalized,
  title = {{Generalized sparse Hadamard single-pixel Imaging}},
  author = {Han, Yuyuan and Wu, Peng and Xu, Hanxiong and Li, Bin and He, Yuchen and Liu, Jianbin and Chen, Hui and Zheng, Huaibin},
  journal = {Optics Express},
  volume = {33},
  number = {20},
  pages = {43324--43341},
  year = {2025},
  doi = {10.1364/OE.576533}
}
  • Low-order matrices, less memory
  • No training or post-processing
  • Two acquisition modes with theory
Complete LDGI OPA figure with the LNOI one-dimensional optical phased-array and grating-waveguide schematic, magnified emission region, and X- and Y-direction far-field curves with all axes and legends.
Fig. 2 · Full size

Loop differential ghost imaging based on line pattern

Yuyuan Han, Huaibin Zheng, Bin Li, Jingwei Li, Long Qiu, Wenxuan Hao, Zheng Dang, Hui Chen, Jianbin Liu, Yuchen He, Yanyan Liu, Zhuo Xu

Optics Communications · 2025

Line patterns from one-dimensional optical phased arrays break the noise symmetry that ghost-imaging denoising relies on; LDGI overcomes this limitation.

PaperCode
BibTeX
@article{han2025loop,
  title = {{Loop differential ghost imaging based on line pattern}},
  author = {Han, Yuyuan and Zheng, Huaibin and Li, Bin and Li, Jingwei and Qiu, Long and Hao, Wenxuan and Dang, Zheng and Chen, Hui and Liu, Jianbin and He, Yuchen and Liu, Yanyan and Xu, Zhuo},
  journal = {Optics Communications},
  volume = {580},
  pages = {131589},
  year = {2025},
  doi = {10.1016/j.optcom.2025.131589}
}
  • Orders-of-magnitude gain over traditional ghost imaging
  • Noise-symmetry metric κs
  • Validated optically

Current directions

  1. Acquisition-aware sensing

    Single-pixel sensing methods that account for acquisition conditions, tested in optical-bench experiments. Manuscript in preparation.

  2. Question answering from measurements

    Exploratory

    Whether a frozen vision–language model can answer questions directly from single-pixel measurements.

  3. Time-domain diffuse optics

    Planned study

    3D reconstruction inside scattering media from photon arrival times, from simulation to phantoms.

Publications

9 papers, newest firstAll BibTeX (.bib)

  1. Graph-Regularized Estimation for Inverse Problems in Computational Optical Imaging

    Chong Wang, Hao Lu, Yuyuan Han, Huaibin Zheng, Yuchen He

    IEEE Transactions on Image Processing · 35 · 9314–9326 · 2026

    Paper
    BibTeX
    @article{wang2026graph,
  title = {{Graph-Regularized Estimation for Inverse Problems in Computational Optical Imaging}},
  author = {Wang, Chong and Lu, Hao and Han, Yuyuan and Zheng, Huaibin and He, Yuchen},
  journal = {IEEE Transactions on Image Processing},
  volume = {35},
  pages = {9314--9326},
  year = {2026},
  doi = {10.1109/TIP.2026.3726365}
}

  2. Optical Phased Arrays on Photonic Integrated Circuit Platforms: Challenges, Opportunities, and Future Applications

    Jingwei Li, Yuyuan Han, Aoxiang Chen, Pengpeng Li, Xinglin Zhu, Guanbao Zhao, Luohan Peng, Huaibin Zheng

    Nanophotonics · 15(14) · e70182 · 2026

    Paper
    BibTeX
    @article{li2026optical,
  title = {{Optical Phased Arrays on Photonic Integrated Circuit Platforms: Challenges, Opportunities, and Future Applications}},
  author = {Li, Jingwei and Han, Yuyuan and Chen, Aoxiang and Li, Pengpeng and Zhu, Xinglin and Zhao, Guanbao and Peng, Luohan and Zheng, Huaibin},
  journal = {Nanophotonics},
  volume = {15},
  number = {14},
  pages = {e70182},
  year = {2026},
  doi = {10.1002/nap2.70182}
}

  3. The Lift Spectrum: How Measurement-to-Space Adaptivity Shapes Robustness in Image-Free Single-Pixel Sensing

    Yuyuan Han, Jingwei Li, Xiaoxia Zhang, Long Qiu, Chong Wang, Wenxuan Hao, Jiangyu Han, Xinyu Yao, Yuchen He, Hui Chen, Jianbin Liu, Huaibin Zheng

    arXiv preprint · 2026

    PaperCode
    BibTeX
    @misc{han2026lift,
  title = {{The Lift Spectrum: How Measurement-to-Space Adaptivity Shapes Robustness in Image-Free Single-Pixel Sensing}},
  author = {Han, Yuyuan and Li, Jingwei and Zhang, Xiaoxia and Qiu, Long and Wang, Chong and Hao, Wenxuan and Han, Jiangyu and Yao, Xinyu and He, Yuchen and Chen, Hui and Liu, Jianbin and Zheng, Huaibin},
  year = {2026},
  eprint = {2607.22077},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2607.22077}
}

  4. Generalized sparse Hadamard single-pixel Imaging

    Yuyuan Han, Peng Wu, Hanxiong Xu, Bin Li, Yuchen He, Jianbin Liu, Hui Chen, Huaibin Zheng

    Optics Express · 33(20) · 43324–43341 · 2025

    PaperCode
    BibTeX
    @article{han2025generalized,
  title = {{Generalized sparse Hadamard single-pixel Imaging}},
  author = {Han, Yuyuan and Wu, Peng and Xu, Hanxiong and Li, Bin and He, Yuchen and Liu, Jianbin and Chen, Hui and Zheng, Huaibin},
  journal = {Optics Express},
  volume = {33},
  number = {20},
  pages = {43324--43341},
  year = {2025},
  doi = {10.1364/OE.576533}
}

  5. Loop differential ghost imaging based on line pattern

    Yuyuan Han, Huaibin Zheng, Bin Li, Jingwei Li, Long Qiu, Wenxuan Hao, Zheng Dang, Hui Chen, Jianbin Liu, Yuchen He, Yanyan Liu, Zhuo Xu

    Optics Communications · 580 · 131589 · 2025

    PaperCode
    BibTeX
    @article{han2025loop,
  title = {{Loop differential ghost imaging based on line pattern}},
  author = {Han, Yuyuan and Zheng, Huaibin and Li, Bin and Li, Jingwei and Qiu, Long and Hao, Wenxuan and Dang, Zheng and Chen, Hui and Liu, Jianbin and He, Yuchen and Liu, Yanyan and Xu, Zhuo},
  journal = {Optics Communications},
  volume = {580},
  pages = {131589},
  year = {2025},
  doi = {10.1016/j.optcom.2025.131589}
}

  6. Design and analysis of single-electrode integrated lithium niobate optical phased array for two-dimensional beam steering

    Jingwei Li, Huaibin Zheng, Yuyuan Han, Bin Li, Wenxuan Hao, Long Qiu, Yanyan Liu, Yuchen He, Xiaoyong Wei, Zhuo Xu

    Optics and Lasers in Engineering · 184 · 108617 · 2025

    Paper
    BibTeX
    @article{li2025design,
  title = {{Design and analysis of single-electrode integrated lithium niobate optical phased array for two-dimensional beam steering}},
  author = {Li, Jingwei and Zheng, Huaibin and Han, Yuyuan and Li, Bin and Hao, Wenxuan and Qiu, Long and Liu, Yanyan and He, Yuchen and Wei, Xiaoyong and Xu, Zhuo},
  journal = {Optics and Lasers in Engineering},
  volume = {184},
  pages = {108617},
  year = {2025},
  doi = {10.1016/j.optlaseng.2024.108617}
}

  7. Fusion of near-infrared single-photon LiDAR with visible light camera

    Long Qiu, Huaibin Zheng, Yuyuan Han, Bin Li, Wenxuan Hao, Zheng Dang, Yuanlin Yu, Jiawei Li, Hanxiong Xu, Hui Chen, Jianbin Liu, Yuchen He, Yanyan Liu, Zhuo Xu

    Measurement Science and Technology · 36(6) · 065409 · 2025

    Paper
    BibTeX
    @article{qiu2025fusion,
  title = {{Fusion of near-infrared single-photon LiDAR with visible light camera}},
  author = {Qiu, Long and Zheng, Huaibin and Han, Yuyuan and Li, Bin and Hao, Wenxuan and Dang, Zheng and Yu, Yuanlin and Li, Jiawei and Xu, Hanxiong and Chen, Hui and Liu, Jianbin and He, Yuchen and Liu, Yanyan and Xu, Zhuo},
  journal = {Measurement Science and Technology},
  volume = {36},
  number = {6},
  pages = {065409},
  year = {2025},
  doi = {10.1088/1361-6501/addddf}
}

  8. Long-Range Imaging LiDAR with Multiple Denoising Technologies

    Huaibin Zheng, Yuyuan Han, Long Qiu, Yanfeng Zong, Jingwei Li, Yu Zhou, Yuchen He, Jianbin Liu, Gao Wang, Hui Chen, Zhuo Xu

    Applied Sciences · 14(8) · 3414 · 2024

    Paper
    BibTeX
    @article{zheng2024long,
  title = {{Long-Range Imaging LiDAR with Multiple Denoising Technologies}},
  author = {Zheng, Huaibin and Han, Yuyuan and Qiu, Long and Zong, Yanfeng and Li, Jingwei and Zhou, Yu and He, Yuchen and Liu, Jianbin and Wang, Gao and Chen, Hui and Xu, Zhuo},
  journal = {Applied Sciences},
  volume = {14},
  number = {8},
  pages = {3414},
  year = {2024},
  doi = {10.3390/app14083414}
}

  9. Sequence-Controlled Pseudothermal Optical Ghost Imaging System

    Yanfeng Zong, Huaibin Zheng, Xinwei Wu, Jingwei Li, Long Qiu, Yuyuan Han

    Acta Optica Sinica · 43(7) · 0711001 · 2023

    Paper
    BibTeX
    @article{zong2023sequence,
  title = {{Sequence-Controlled Pseudothermal Optical Ghost Imaging System}},
  author = {Zong, Yanfeng and Zheng, Huaibin and Wu, Xinwei and Li, Jingwei and Qiu, Long and Han, Yuyuan},
  journal = {Acta Optica Sinica},
  volume = {43},
  number = {7},
  pages = {0711001},
  year = {2023},
  doi = {10.3788/AOS221781}
}

Research code

  • STSF-TPLS

    Image-free segmentation from single-pixel measurements with a content-adaptive lift and task-prioritized loss scheduling; code and pretrained weights (PyTorch).

  • Comparison-of-SPI-DLs

    Eight deep architectures and a classical inverse, benchmarked on one shared Hadamard acquisition (PyTorch).

  • SHSI

    Sparse Hadamard single-pixel imaging: pattern generation and reconstruction (MATLAB).

  • LDGI

    Loop differential ghost imaging for line patterns from optical phased arrays (MATLAB).

  • TGIs-Entry

    Reference implementations of traditional, differential and normalized ghost imaging (MATLAB).

  • remote-gpu-trainer

    Open-source agent skill for Claude Code and Codex that runs, verifies and delivers GPU experiments (60+ stars).

Background

Education

  • Xi'an Jiaotong University

    PhD student · Electronic Science and Technology · Master’s-to-PhD track

    Advisor: Huaibin Zheng

    2022–present

  • Politecnico di Milano

    Visiting PhD student

    Department of Physics

    Supervisor: Antonio Pifferi

    Oct 2026–present

  • Northwest University

    B.Eng. · Electronic Information Science and Technology

    2018–2022

Additional experience

Photonic device modeling · Modeled and optimized electro-optic phased-array geometries for beam steering.

Optical system integration · Led mechanical design, optical alignment and system integration; participated in underwater field tests.

Award

  • Second Prize, National Post-Graduate Mathematical Contest in Modeling2024 Code

Service

  • Reviewer for IEEE Transactions on Image Processing, IEEE Transactions on Computational Imaging, Optics Express and Optics Letters13 verified reviews · 2025–2026

Contact

hanyuyuan6@gmail.com