Identification of image global processing operator chain based on feature decoupling

doi:10.1016/j.ins.2023.118961

CORC > 自动化研究所 > 中国科学院自动化研究所 > 多模态人工智能系统全国重点实验室

	Identification of image global processing operator chain based on feature decoupling
	Chen, Jiaxin 1; Liao, Xin 1; Wang, Wei 2; Qin, Zheng 1
刊名	INFORMATION SCIENCES
	2023-08-01
卷号	637 页码:18
关键词	Image forensics Global processing operator chain Operator chain identification Feature decoupling
ISSN号	0020-0255
DOI	10.1016/j.ins.2023.118961
通讯作者	Liao, Xin(xinliao@hnu.edu.cn)
英文摘要	Image authenticity verification is an important issue to be studied, which has attracted growing attention recently. Most of the existing forensic methods are aimed at detecting a specific manipulation. However, due to the superimposed processing artifacts caused by using different operations to forge images, the image global processing operator chain identification, which is composed of multiple global manipulations in a certain order, remains a challenge. In this paper, we focus on detecting multiple manipulations and identifying the order of these manipulations. By analyzing the relationship between blind signal separation and operator chain identification, we find that the independent source features of different operations will be coupled when the image is processed by multiple operations, which is similar to what in blind signal separation. Therefore, it is reasonable to formulate the problem of operator chain identification with blind signal separation. Then, a feature decoupling method is proposed to estimate the source feature from the coupled features by optimizing a decoupling matrix. These estimated decoupled features are valid evidence for operator chain identification. For the realistic scenario where images are saved in JPEG format, the comparison with some state-of-the-art methods demonstrates that the proposed method could identify operator chains with better performance.
资助项目	National Natural Science Foundation of China[U22A2030] ; National Natural Science Foundation of China[U20A20174] ; National Natural Science Foundation of China[61972142] ; National Natural Science Foundation of China[61972395] ; National Natural Science Foundation of China[62002112] ; Natural Science Foundation of Hunan Province[2020JJ4212] ; Natural Science Foundation of Hunan Province[2021JJ40117]
WOS关键词	FORENSICS
WOS研究方向	Computer Science
语种	英语
出版者	ELSEVIER SCIENCE INC
WOS记录号	WOS:000983689200001
资助机构	National Natural Science Foundation of China ; Natural Science Foundation of Hunan Province
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/53284]
专题	多模态人工智能系统全国重点实验室
通讯作者	Liao, Xin
作者单位	1.Hunan Univ, Comp Sci & Elect Engn, Changsha 410082, Peoples R China 2.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Chen, Jiaxin,Liao, Xin,Wang, Wei,et al. Identification of image global processing operator chain based on feature decoupling[J]. INFORMATION SCIENCES,2023,637:18.
APA	Chen, Jiaxin,Liao, Xin,Wang, Wei,&Qin, Zheng.(2023).Identification of image global processing operator chain based on feature decoupling.INFORMATION SCIENCES,637,18.
MLA	Chen, Jiaxin,et al."Identification of image global processing operator chain based on feature decoupling".INFORMATION SCIENCES 637(2023):18.