Proceedings of the 6th ACM Workshop on Information Hiding and Multimedia Security 2018
DOI: 10.1145/3206004.3206010
|View full text |Cite
|
Sign up to set email alerts
|

Image Forgery Localization based on Multi-Scale Convolutional Neural Networks

Abstract: In this paper, we propose to utilize Convolutional Neural Networks (CNNs) and the segmentation-based multi-scale analysis to locate tampered areas in digital images. First, to deal with color input sliding windows of different scales, a unified CNN architecture is designed. Then, we elaborately design the training procedures of CNNs on sampled training patches. With a set of robust multi-scale tampering detectors based on CNNs, complementary tampering possibility maps can be generated. Last but not least, a se… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
62
0

Year Published

2019
2019
2022
2022

Publication Types

Select...
7
3

Relationship

0
10

Authors

Journals

citations
Cited by 89 publications
(62 citation statements)
references
References 25 publications
(61 reference statements)
0
62
0
Order By: Relevance
“…Interestingly, the first proposed architectures, inspired again by work in steganalysis [47], all focus on suppressing the scene content, forcing the network to work on noise residuals. This is obtained by adding a first layer of highpass filters, either fixed [48], [49], or trainable [50], or else by recasting a conventional feature extractor as a convolutional neural network (CNN) [51]. A two-stream network is proposed in [52], [53] to exploit both low-level and high-level features, where a first network constrained to work on noise residuals is joined with a general purpose deep CNN (ResNet 101 in [53]).…”
Section: B Using Deep Learning For Image Forensicsmentioning
confidence: 99%
“…Interestingly, the first proposed architectures, inspired again by work in steganalysis [47], all focus on suppressing the scene content, forcing the network to work on noise residuals. This is obtained by adding a first layer of highpass filters, either fixed [48], [49], or trainable [50], or else by recasting a conventional feature extractor as a convolutional neural network (CNN) [51]. A two-stream network is proposed in [52], [53] to exploit both low-level and high-level features, where a first network constrained to work on noise residuals is joined with a general purpose deep CNN (ResNet 101 in [53]).…”
Section: B Using Deep Learning For Image Forensicsmentioning
confidence: 99%
“…However, only few works [7], [14] attempt to localize image manipulation at pixel level. Some recent works [17], [26], [54] address the localization problem by classifying patches as manipulated. The localization of image tampering is a very challenging task as well-manipulated images do not leave any visual clues, as shown by the following examples in Fig.…”
Section: Introductionmentioning
confidence: 99%
“…Furthermore, some methods were proposed in the literature [Cozzolino and Verdoliva (2016); Bondi, Lameri, Güera et al (2017)] to solve the image splicing problem, but they were based on some certain assumptions and thus greatly reduced the general applicability of the algorithms. Moreover, the literature [Rao and Ni (2016); Liu, Guan, Zhao et al (2017)] has proposed an identification method for image splicing, but it did not realize the location of the tampering region. Moreover, the literature Zhang et al [Zhang, Goh, Win et al (2016)] proposed a two-stage deep learning method to learn the image corresponding block features to detect tampered images of different image formats.…”
Section: Survey Of Previous Related Workmentioning
confidence: 99%