Jinshan Pan

Also known as Jin-shan Pan, Pan Jinshan, 潘金山

Jinshan Pan is a computer vision researcher whose work on restoring blurred images spans the dark channel prior, dual convolutional networks for low-level vision and frequency-domain Transformers. During his doctoral studies at Dalian University of Technology, he worked with Ming-Hsuan Yang at UC Merced and visited Harvard, and he presented the dark channel prior for blind deblurring as an oral paper at CVPR 2016. At CVPR 2018 he co-authored the DualCNN and a conditional GAN for single-image dehazing, and in 2019 he co-authored the stereo deblurring network DAVANet. A frequency-domain Transformer for deblurring that he co-authored appeared at CVPR 2023 as a Highlight, and he became an Associate Editor of IEEE TPAMI in June 2024. He is a professor at Nanjing University of Science and Technology, where his research covers image deblurring, image and video enhancement, and related vision problems.

Now at: Nanjing University of Science and Technology

Timeline

2012–2013

Kernel estimation from salient structure

In December 2012, he and Risheng Liu, Zhixun Su and Xianfeng Gu post Kernel Estimation from Salient Structure for Robust Motion Deblurring, which estimates a blur kernel from a single blurred image by computing image structure with a total-variation model. The paper is accepted by Signal Processing: Image Communication in 2013.
2014–2016

Doctoral work at Merced and Harvard

As a PhD student in the School of Mathematical Sciences at Dalian University of Technology, he spends September 2014 to April 2016 as a joint-training PhD student in electrical engineering and computer science at UC Merced, directed by Ming-Hsuan Yang. In 2015, he is a visiting PhD student in the Visual Computing Group at Harvard, working with Hanspeter Pfister and Deqing Sun.
2016–2018

The dark channel prior

At CVPR 2016, he presents Blind Image Deblurring Using Dark Channel Prior with Deqing Sun, Hanspeter Pfister and Ming-Hsuan Yang as an oral paper. The method rests on the observation that the dark channel of blurred images is less sparse, and a TPAMI extension, Deblurring Images via Dark Channel Prior, follows in 2018.
2018

Six papers at CVPR 2018

In March 2018, six of his papers, one of them a Spotlight, are accepted at CVPR 2018. They include Learning Dual Convolutional Neural Networks for Low-Level Vision, whose DualCNN has one branch for structures and one for details, and Single Image Dehazing via Conditional Generative Adversarial Network, written with Runde Li, Zechao Li and Jinhui Tang.
2019

DAVANet and stereo deblurring

In 2019, he co-authors DAVANet, a stereo deblurring network with depth awareness and view aggregation, which is an oral presentation at CVPR 2019. The paper comes with a dataset of 20,637 blurry-sharp stereo image pairs from 135 sequences.
2022–2023

A frequency-domain deblurring Transformer

In November 2022, he and co-authors Lingshun Kong, Jiangxin Dong, Mingqiang Li and Jianjun Ge post Efficient Frequency Domain-based Transformers for High-Quality Image Deblurring, which estimates self-attention with an element-wise product in the frequency domain. The paper appears at CVPR 2023 as a Highlight, and in March 2023 four of his papers, two of them highlights, are accepted at the conference.
2023–2024

Editor and area chair roles

In March 2023, he is invited to be an Area Editor for CVIU, and in June 2024 an Associate Editor for IEEE TPAMI. He is also invited as an Area Chair for CVPR 2023, ICLR 2024, ICML 2024 and NeurIPS 2024.
As of October 2026

Professor at NJUST

He is a professor at Nanjing University of Science and Technology, where his research covers image deblurring, image and video enhancement and analysis, and related vision problems.