Ruikang Li

I am a second-year Ph.D. student at MMLab, The Chinese University of Hong Kong, advised by Prof. Tianfan Xue.

Prior to that, I majored in Applied Optics and obtained my Msc degree from Zhejiang University, where I worked closely with Dr. Shiqi Chen, and received my BSc degree in Opto-Electronics from Tianjin University.

Google Scholar  /  Github
Email: lyricccco44 at gmail dot com

I work on Computational Photography and Generative Models for efficient, high-fidelity image generation. My previous work also includes End-to-end Optimization of ISPs (Image Signal Processing) and Opto-physical Simulations.

(*: Equal Contribution †: Corresponding Author)

We propose to jointly optimize denoisers in the RAW and sRGB domains bridged by a dedicated differentiable ISP. Once trained with synthetic noise, it can work on real noise for unseen smartphones and public benchmark.

We propose a multi-stage model for progressive low-light image restoration, disentangling illumination enhancement and noise removal within each stage to mitigate ISP-induced coupling.

We present a physics-based optical imaging-chain simulation for remote sensing that models the full pipeline from 3D surface scattering to the camera's DV output via 3D structures, spectral properties, and calibration.

We build a physics-based model tailored to the challenging illumination conditions of permanently shadowed regions, integrating terrain geometry, orbital configuration, camera parameters, and environmental settings.