Haishan Liu

Reconstructing objects hidden behind complex scattering media remains a fundamental challenge in optical imaging. Existing deep learning methods often require extensive training data and generalize poorly in varying scattering conditions. To address these limitations, we introduce an adapter-enhanced diffusion model (ADM) that combines the powerful denoising capability of diffusion models with a test-time adapter, allowing robust cross-domain alignment even with limited training data. Trained on only 100 paired images from a single indoor scattering condition, ADM surpasses classical Retinex optimization, U-Net, SwinUNet, DescatterNet, and a diffusion model baseline, consistently achieving superior image quality across diverse scenarios—from higher scatterer concentrations to different media types, including artificial fog, milk, and real outdoor fog. ADM demonstrates remarkable data efficiency, performing comparably to U-Net models trained with 12 times more data. Attention map analyses reveal that the test-time adapter effectively steers the diffusion process toward semantically meaningful regions, supporting ADM’s robust generalization from limited data. Our results highlight ADM as a promising and efficient solution for imaging under real-world scattering conditions. Code and pre-trained models will be released for reproducibility.

It is widely recognized that optically captured images can be distorted under non-ideal imaging conditions, such as blurring, defocusing, and the presence of scattering media. In this study, we introduce a novel architecture that integrates an adaptive encoder to enhance the general restoration of optical images. We refer to this architecture as AdaptiveNet, which is capable of adaptively identifying different types of non-ideal imaging conditions and learning to restore images accordingly. Our extensive experimental results indicate that AdaptiveNet outperforms existing methods across a range of optical image restoration tasks.

Imaging through dynamic scattering media is one of the most challenging yet fascinating problems in optics, with applications spanning from biological detection to remote sensing. In this study, we propose a comprehensive learning-based technique that facilitates real-time, non-invasive, incoherent imaging of real-world objects through dense and dynamic scattering media. We conduct extensive experiments, demonstrating the capability of our technique to see through turbid water and natural fog. The experimental results indicate that the proposed technique surpasses existing approaches in numerous aspects and holds significant potential for imaging applications across a broad spectrum of disciplines.

Imaging through non-static and optically thick scattering media such as dense fog, heavy smoke, and turbid water is crucial in various applications. However, most existing methods rely on either active and coherent light illumination, or image priors, preventing their application in situations where only passive illumination is possible. In this study we present a universal passive method for imaging through dense scattering media that does not depend on any prior information. Combining the selection of small-angle components out of the incoming information-carrying scattering light and image enhancement algorithm that incorporates time-domain minimum filtering and denoising, we show that the proposed method can dramatically improve the signal-to-interference ratio and contrast of the raw camera image in outfield experiments.

We study the potential LHC discovery of the Z1 KK gauge boson unitarizing longitudinal W+W- scattering amplitude. In particular, we explore the decay mode Z1->t tbar along with Z1-> W+W- without specifying the branching fractions. We propose to exploit the associated production pp-> W Z1, and select the final state of like-sign dileptons plus multijets and large missing energy. We conclude that it is possible to observe the Z1 resonance at a 5 sigma level with an integrated luminosity of 100 inverse fb at the LHC upto 650 GeV for a dominant WW channel, and 560 GeV for a dominant ttbar channel.Comment: 13 pages, 7 figure