Lin Liu
Currently, I am a joint PhD student at University of Science and Technology of China (USTC) and Shanghai AI Laboratory. I was a master student in the School of Information of Science and Technology, University of Science and Technology of China(USTC). Previously, I spent 8 months at the Noah's Ark Lab, Huawei Tech as a research intern, working with Prof. Jianzhuang Liu. I work closely with Shanxin Yuan, Xu Jia, Lingxi Xie, Mingming Zhao, and Prof. Qi Tian.
Education
News
Projects or Competitions
Talks
Abstract: In recent years, with the development of deep learning, image restoration and enhancement, such as image denoising and image super-resolution, have attracted more and more attention. This talk will focus on the challenges faced by image restoration and propose some solutions: 1) For the problem that real data is difficult to obtain, we introduce methods of synthesizing more real data, self-supervised learning, and fine-tuning using pre-trained models. 2) For the existing models have limited in mining useful information, we first introduce the concept of ‘guidance restoration’, then introduce some self-guidance and external-guidance methods. 3) At last, we introduce some video or burst methods for image restoration.
Publications
I'm interested in low-level vision, computational photography, image/video compression, and Neural radiance fields (NeRF). Currently, I have four CCF-A conferences papers, two CCF-B conferences (ECCV) papers, and two Transactions journal papers.
Low-Light Video Enhancement with Synthetic Event Guidance, AAAI 2023. [PDF]
TAPE: Task-Agnostic Prior Embedding for Image Restoration, ECCV 2022. [PDF]
SiamTrans: Zero-Shot Multi-Frame Image Restoration with Pre-Trained Siamese Transformers. AAAI 2022.
Self-Adaptively Learning to Demoiré from Focused and Defocused Image Pairs, NIPS 2020 .[PDF]
Wavelet-Based Dual-Branch Networkfor Image Demoireing, ECCV 2020 . [PDF]
Joint Demosaicing and Denoising with Self Guidance, CVPR 2020 . [PDF]
Learning Frequency Domain Priors for Image Demoireing. TPAMI 2021.
HRInversion: High-Resolution GAN Inversion for Cross-Domain Image Synthesis. TCSVT 2022.
Exploring Effective Mask Sampling Modeling for Neural Image Compression. Arxiv 2023. [PDF]
Joint Channel Estimation and Feedback with Masked Token Transformers in Massive MIMO Systems. Arxiv 2023. [PDF]
Conference Papers:
Low-Light Video Enhancement with Synthetic Event Guidance
(AAAI), 2023
Lin Liu, Junfeng An, Jianzhuang Liu, Shanxin Yuan, Xiangyu Chen, Wengang Zhou, Houqiang Li, Yanfeng Wang, Qi Tian
In this paper, inspired by the low latency and high dynamic range of events, we use synthetic events from multiple frames to guide the enhancement and restoration of low-light videos. Our method contains three stages.
PDF (Arxiv) / Project Page (Comming soon!) / Code (MindSpore) /
TAPE: Task-Agnostic Prior Embedding for Image Restoration
(ECCV), 2022
Lin Liu, Lingxi Xie, Xiaopeng Zhang, Shanxin Yuan, Xiangyu Chen, Wengang Zhou, Houqiang Li, Qi Tian
In this paper, we propose a novel approach that embeds a task-agnostic prior into a transformer. Our approach, named Task-Agnostic Prior Embedding (TAPE), consists of three stages, namely, task-agnostic pre-training, task-agnostic fine-tuning, and task-specific fine-tuning, where the first one embeds prior knowledge about natural images into the transformer and the latter two extracts the knowledge to assist downstream image restoration.
PDF / Project Page / Code /
SiamTrans: Zero-Shot Multi-Frame Image Restoration with Pre-Trained Siamese Transformers
(AAAI), 2022
Lin Liu, Shanxin Yuan, Jianzhuang Liu, Xin Guo, Youliang Yan, Qi Tian
We propose a novel zero-shot multi-frame image restoration method for removing unwanted obstruction elements (such as rains, snow, and moire patterns) that vary in successive frames.
PDF / Codes (Release soon! ) / Project page
Self-Adaptively Learning to Demoiré from Focused and Defocused Image Pairs
Thirty-fourth Conference on Neural Information Processing Systems(NeurIPS), 2020
Lin Liu, Shanxin Yuan, Jianzhuang Liu, Liping Bao, Gregory Slabaugh, Qi Tian
We propose a self-adaptive learning method for image demoiréing, which uses an additional input (defocused moiré-free image) to help remove moiré patterns from the focused moiré image.
PDF / Codes / Project page
Wavelet-Based Dual-Branch Networkfor Image Demoireing
European Conference on Computer Vision(ECCV), 2020
Lin Liu, Jianzhuang Liu, Shanxin Yuan, Gregory Slabaugh, Ales Leonardis, Wengang Zhou, Qi Tian
We propose a novel wavelet-based and dual-branch neural network for image demoireing. Our network removes moire patterns in the wavelet domain to separate the frequencies of moire patterns from the image content.
PDF / Codes
Joint Demosaicing and Denoising with Self Guidance
Conference on Computer Vision and Pattern Recognition(CVPR), 2020
Lin Liu, Xu Jia, Jianzhuang Liu, Qi Tian
We propose a novel self-guidance network (SGNet) with density-map guidance and green-channel guidance for joint demosaicing and denoising (JDD). And it works well in super-resolution too.
PDF / Codes
Journal Papers:
Video Demoireing with Deep Temporal Color Embedding and Video-Image Invertible Consistency
IEEE Transactions on Multimedia, 2023
Lin Liu, Junfeng An, Shanxin Yuan, Wengang Zhou , Houqiang Li, Yanfeng Wang, Qi Tian
In this paper, we present a novel deep learning-based approach called the Deep Temporal Color Embedding network (DTCENet) that employs an invertible network to align distortion color patches in a patch-based embedding framework for video demoireing.
PDF (coming soon) / Codes (coming soon) /
Learning Frequency Domain Priors for Image Demoireing
IEEE Transactions on Pattern Analysis and Machine Intelligence ( TPAMI ), 2021
Bolun Zheng; Shanxin Yuan; Chenggang Yan; Xiang Tian; Jiyong Zhang; Yaoqi Sun; Lin Liu ; Ales Leonardis; Greg Slabaugh
In this paper, we raise a general degradation model to describe an image contaminated by moire patterns, and propose a novel multi-scale bandpass convolutional neural network (MBCNN) for single image demoireing.
PDF / Codes /
HRInversion: High-Resolution GAN Inversion for Cross-Domain Image Synthesis
( TCSVT ), 2022
Peng Zhou; Lingxi Xie; BingBing Ni; Lin Liu ; Qi Tian
In this paper, we use the GAN inversion method for cross-domain image synthesis.
PDF / Codes /
Awards
2022.10 Suzhou Yucai Scholarship (苏州育才奖学金)
2019.09 National Network Security Scholarship (国家网络安全奖学金) (link)
2018.09 National Scholarship (国家奖学金) (highest national wide scholarship for students in China)
2019.09 Outstanding Graduate Student of Anhui Province (安徽省优秀毕业生)
2017.09 Scholarship of Institute of electronics, Chinese Academy of Sciences (中国科学院电子所奖学金)
2017.10 中国科学技术大学机器人竞赛，第二名