MIPI 2024 挑战：混合EVS 相机去彩色化方法与结果
图像与视频处理
2024-05-09 v1 计算机视觉与模式识别
摘要
对移动平台的计算摄影和成像需求日益增长，推动了高级图像传感器与新型算法在相机系统中的广泛开发与集成。然而，缺乏高质量数据以及学术界与工业界交流的罕见机会，限制了移动智能摄影和成像（MIPI）的发展。建立在 ECCV 2022 和 CVPR 2023 之前 MIPI 工作坊基础上，我们介绍了本次第三届 MIPI 挑战赛，包括三个聚焦 novel image sensors 和 imaging algorithms 的轨道。本文总结和回顾了 MIPI 2024 中夜间晕影消除轨道的结果。总共成功注册了 170 名参赛者，14 支队伍在最终测试阶段提交了结果。本次挑战中开发出的解决方案在夜间晕影消除方面实现了最先进的性能。挑战的更多细节及数据集链接可在 https://mipi-challenge.org/MIPI2024/ 查找。
引用
@article{arxiv.2405.04867,
title = {MIPI 2024 Challenge on Demosaic for HybridEVS Camera: Methods and Results},
author = {Yaqi Wu and Zhihao Fan and Xiaofeng Chu and Jimmy S. Ren and Xiaoming Li and Zongsheng Yue and Chongyi Li and Shangcheng Zhou and Ruicheng Feng and Yuekun Dai and Peiqing Yang and Chen Change Loy and Senyan Xu and Zhijing Sun and Jiaying Zhu and Yurui Zhu and Xueyang Fu and Zheng-Jun Zha and Jun Cao and Cheng Li and Shu Chen and Liang Ma and Shiyang Zhou and Haijin Zeng and Kai Feng and Yongyong Chen and Jingyong Su and Xianyu Guan and Hongyuan Yu and Cheng Wan and Jiamin Lin and Binnan Han and Yajun Zou and Zhuoyuan Wu and Yuan Huang and Yongsheng Yu and Daoan Zhang and Jizhe Li and Xuanwu Yin and Kunlong Zuo and Yunfan Lu and Yijie Xu and Wenzong Ma and Weiyu Guo and Hui Xiong and Wei Yu and Bingchun Luo and Sabari Nathan and Priya Kansal},
journal= {arXiv preprint arXiv:2405.04867},
year = {2024}
}
备注
MIPI@CVPR2024. Website: https://mipi-challenge.org/MIPI2024/