NTIRE 2023 Challenge on 360deg Omnidirectional Image and Video Super-Resolution: Datasets, Methods and Results

Mingdeng Cao, Chong Mou, Fanghua Yu, Xintao Wang, Yinqiang Zheng, Jian Zhang, Chao Dong, Gen Li, Ying Shan, Radu Timofte, Xiaopeng Sun, Weiqi Li, Zhenyu Zhang, Xuhan Sheng, Bin Chen, Haoyu Ma, Ming Cheng, Shijie Zhao, Wanwan Cui, Tianyu Xu, Chunyang Li, Long Bao, Heng Sun, Huaibo Huang, Xiaoqiang Zhou, Yuang Ai, Ran He, Renlong Wu, Yi Yang, Zhilu Zhang, Shuohao Zhang, Junyi Li, Yunjin Chen, Dongwei Ren, Wangmeng Zuo, Qian Wang, Hao-Hsiang Yang, Yi-Chung Chen, Zhi-Kai Huang, Wei-Ting Chen, Yuan-Chun Chiang, Hua-En Chang, I-Hsiang Chen, Chia-Hsuan Hsieh, Sy-Yen Kuo, Zebin Zhang, Jiaqi Zhang, Yuhui Wang, Shuhao Cui, Junshi Huang, Li Zhu, Shuman Tian, Wei Yu, Bingchun Luo; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023, pp. 1731-1745

Abstract

This report introduces two high-quality datasets Flickr360 and ODV360 for omnidirectional image and video super-resolution, respectively, and reports the NTIRE 2023 challenge on 360deg omnidirectional image and video super-resolution. Unlike ordinary 2D images/videos with a narrow field of view, omnidirectional images/videos can represent the whole scene from all directions in one shot. There exists a large gap between omnidirectional image/video and ordinary 2D image/video in both the degradation and restoration processes. The challenge is held to facilitate the development of omnidirectional image/video super-resolution by considering their special characteristics. In this challenge, two tracks are provided: one is the omnidirectional image super-resolution and the other is the omnidirectional video super-resolution. The task of the challenge is to super-resolve an input omnidirectional image/video with a magnification factor of x4. Realistic omnidirectional downsampling is applied to construct the datasets. Some general degradation(e.g., video compression) is also considered for the video track. The challenge has 100 and 56 registered participants for those two tracks. In the final testing stage, 7 and 3 participating teams submitted their results, source codes, and fact sheets. Almost all teams achieved better performance than baseline models by integrating omnidirectional characteristics, reaching compelling performance on our newly collected Flickr360 and ODV360 datasets.

Related Material

[pdf]
[bibtex]
@InProceedings{Cao_2023_CVPR, author = {Cao, Mingdeng and Mou, Chong and Yu, Fanghua and Wang, Xintao and Zheng, Yinqiang and Zhang, Jian and Dong, Chao and Li, Gen and Shan, Ying and Timofte, Radu and Sun, Xiaopeng and Li, Weiqi and Zhang, Zhenyu and Sheng, Xuhan and Chen, Bin and Ma, Haoyu and Cheng, Ming and Zhao, Shijie and Cui, Wanwan and Xu, Tianyu and Li, Chunyang and Bao, Long and Sun, Heng and Huang, Huaibo and Zhou, Xiaoqiang and Ai, Yuang and He, Ran and Wu, Renlong and Yang, Yi and Zhang, Zhilu and Zhang, Shuohao and Li, Junyi and Chen, Yunjin and Ren, Dongwei and Zuo, Wangmeng and Wang, Qian and Yang, Hao-Hsiang and Chen, Yi-Chung and Huang, Zhi-Kai and Chen, Wei-Ting and Chiang, Yuan-Chun and Chang, Hua-En and Chen, I-Hsiang and Hsieh, Chia-Hsuan and Kuo, Sy-Yen and Zhang, Zebin and Zhang, Jiaqi and Wang, Yuhui and Cui, Shuhao and Huang, Junshi and Zhu, Li and Tian, Shuman and Yu, Wei and Luo, Bingchun}, title = {NTIRE 2023 Challenge on 360deg Omnidirectional Image and Video Super-Resolution: Datasets, Methods and Results}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops}, month = {June}, year = {2023}, pages = {1731-1745} }