-
[pdf]
[bibtex]@InProceedings{Zhang_2023_CVPR, author = {Zhang, Yulun and Zhang, Kai and Chen, Zheng and Li, Yawei and Timofte, Radu and Zhang, Junpei and Zhang, Kexin and Peng, Rui and Ma, Yanbiao and Jia, Licheng and Huang, Huaibo and Zhou, Xiaoqiang and Ai, Yuang and He, Ran and Qiu, Yajun and Zhu, Qiang and Li, Pengfei and Li, Qianhui and Zhu, Shuyuan and Zhang, Dafeng and Li, Jia and Wang, Fan and Li, Chunmiao and Kim, TaeHyung and Kil, Jungkeong and Kim, Eon and Yu, Yeonseung and Lee, Beomyeol and Lee, Subin and Lim, Seokjae and Chae, Somi and Choi, Heungjun and Huang, ZhiKai and Chen, YiChung and Chiang, YuanChun and Yang, HaoHsiang and Chen, WeiTing and Chang, HuaEn and Chen, I-Hsiang and Hsieh, ChiaHsuan and Kuo, SyYen and Choi, Ui-Jin and Conde, Marcos V. and Khowaja, Sunder Ali and Yoon, Jiseok and Lee, Ik Hyun and Gendy, Garas and Sabor, Nabil and Hou, Jingchao and He, Guanghui and Zhang, Zhao and Li, Baiang and Zheng, Huan and Zhao, Suiyi and Gao, Yangcheng and Wei, Yanyan and Ren, Jiahuan and Wei, Jiayu and Li, Yanfeng and Sun, Jia and Cheng, Zhanyi and Li, Zhiyuan and Yao, Xu and Wang, Xinyi and Li, Danxu and Cui, Xuan and Cao, Jun and Li, Cheng and Zheng, Jianbin and Sarvaiya, Anjali and Prajapati, Kalpesh and Patra, Ratnadeep and Barik, Pragnesh and Rathod, Chaitanya and Upla, Kishor and Raja, Kiran and Ramachandra, Raghavendra and Busch, Christoph}, title = {NTIRE 2023 Challenge on Image Super-Resolution (x4): Methods and Results}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, month = {June}, year = {2023}, pages = {1865-1884} }
NTIRE 2023 Challenge on Image Super-Resolution (x4): Methods and Results
Abstract
This paper reviews the NTIRE 2023 challenge on image super-resolution (x4), focusing on the proposed solutions and results. The task of image super-resolution (SR) is to generate a high-resolution (HR) output from a corresponding low-resolution (LR) input by leveraging prior information from paired LR-HR images. The aim of the challenge is to obtain a network design/solution capable to produce high-quality results with the best performance (e.g., PSNR). We want to explore how high performance we can achieve regardless of computational cost (e.g., model size and FLOPs) and data. The track of the challenge was to measure the restored HR images with the ground truth HR images on DIV2K testing dataset. The ranking of the teams is determined directly by the PSNR value. The challenge has attracted 192 registered participants, where 15 teams made valid submissions. They achieve state-of-the-art performance in single image super-resolution.
Related Material