Multi-Cue Semi-Supervised Color Constancy With Limited Training Samples

doi:10.1109/TIP.2020.3007823

CORC > 自动化研究所 > 中国科学院自动化研究所 > 模式识别国家重点实验室 > 视频内容安全团队

	Multi-Cue Semi-Supervised Color Constancy With Limited Training Samples
	Huang, Xinwei 7; Li, Bing 2,6; Li, Shuai 7; Li, Wenjuan 2,6; Xiong, Weihua 2,6; Yin, Xuanwu 5; Hu, Weiming 1,4; Qin, Hong 3
刊名	IEEE TRANSACTIONS ON IMAGE PROCESSING
	2020
卷号	29 页码:7875-7888
关键词	Color constancy illumination estimation white balancing multi-cue semi-supervised
ISSN号	1057-7149
DOI	10.1109/TIP.2020.3007823
通讯作者	Li, Bing(bli@nlpr.ia.ac.cn)
英文摘要	Color constancy is one of the fundamental tasks in computer vision. Many supervised methods, including recently proposed Convolutional Neural Networks (CNN)-based methods, have been proved to work well on this problem, but they often require a sufficient number of labeled data. However, it is expensive and time-consuming to collect a large number of labeled training images with accurately measured illumination. In order to reduce the dependence on labeled images and leverage unlabeled ones without measured illumination, we propose a novel semi-supervised framework with limited training samples for illumination estimation. Our key insight is that the images with similar features from different cues will share similar lighting conditions. Consequently, three graphs based on three visual cues, low-level RGB color distribution, mid-level initial illuminant estimates and high-level scene content, are constructed to represent the relationship among different images. Then a multi-cue semi-supervised color constancy method (MSCC) is proposed after integrating these three graphs into a unified model. Extensive experiments on benchmark datasets demonstrate that our proposed MSCC method outperforms nearly all the existing supervised methods with limited labeled samples. Even with no unlabeled samples, MSCC still obtains better performance and stableness than most supervised methods.
资助项目	Beijing Natural Science Foundation[JQ18018] ; Beijing Natural Science Foundation[L172051] ; National Key Research and Development Program of China[2017YFB1002801] ; National Key Research and Development Program of China[2016QY01W0106] ; National Key Research and Development Program of China[2017YFF0106407] ; National Natural Science Foundation of China[U1936204] ; National Natural Science Foundation of China[U1803119] ; National Natural Science Foundation of China[U1736106] ; National Natural Science Foundation of China[61532002] ; National Natural Science Foundation of China[61906192] ; National Natural Science Foundation of China[61876100] ; USA NSF[IIS-1715985] ; USA NSF[1812606] ; Youth Innovation Promotion Association, CAS
WOS关键词	ILLUMINATION CHROMATICITY ; ALGORITHMS
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:000552264300004
资助机构	Beijing Natural Science Foundation ; National Key Research and Development Program of China ; National Natural Science Foundation of China ; USA NSF ; Youth Innovation Promotion Association, CAS
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/40209]
专题	自动化研究所_模式识别国家重点实验室_视频内容安全团队
通讯作者	Li, Bing
作者单位	1.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China 2.Inst Automat, Natl Lab Pattern Recognit NLPR, Beijing 100190, Peoples R China 3.SUNY Stony Brook, Dept Comp Sci, Stony Brook, NY 11794 USA 4.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, CAS Ctr Excellence Brain Sci & Intelligence Techn, Beijing 100190, Peoples R China 5.Hisilicon, Dept Kirin Chipset & Technol Dev, Beijing 100095, Peoples R China 6.PeopleAI Inc, Beijing 100190, Peoples R China 7.Beihang Univ, State Key Lab Virtual Real Technol & Syst, Beijing 100191, Peoples R China
推荐引用方式 GB/T 7714	Huang, Xinwei,Li, Bing,Li, Shuai,et al. Multi-Cue Semi-Supervised Color Constancy With Limited Training Samples[J]. IEEE TRANSACTIONS ON IMAGE PROCESSING,2020,29:7875-7888.
APA	Huang, Xinwei.,Li, Bing.,Li, Shuai.,Li, Wenjuan.,Xiong, Weihua.,...&Qin, Hong.(2020).Multi-Cue Semi-Supervised Color Constancy With Limited Training Samples.IEEE TRANSACTIONS ON IMAGE PROCESSING,29,7875-7888.
MLA	Huang, Xinwei,et al."Multi-Cue Semi-Supervised Color Constancy With Limited Training Samples".IEEE TRANSACTIONS ON IMAGE PROCESSING 29(2020):7875-7888.