2018
DOI: 10.48550/arxiv.1801.06510
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

Image Provenance Analysis at Scale

Abstract: Prior art has shown it is possible to estimate, through image processing and computer vision techniques, the types and parameters of transformations that have been applied to the content of individual images to obtain new images. Given a large corpus of images and a query image, an interesting further step is to retrieve the set of original images whose content is present in the query image, as well as the detailed sequences of transformations that yield the query image given the original images. This is a pro… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

1
18
0

Year Published

2018
2018
2019
2019

Publication Types

Select...
2
1

Relationship

1
2

Authors

Journals

citations
Cited by 3 publications
(19 citation statements)
references
References 40 publications
(70 reference statements)
1
18
0
Order By: Relevance
“…In order to overcome the direction limitation and propose a scalable approach, a more complete end-to-end pipeline for image provenance analysis was described in [52]. That method for graph construction first builds dissimilarity matrices based on local image features, and then employs hierarchical clustering to group nodes and draw edges within the final provenance graph.…”
Section: Related Workmentioning
confidence: 99%
See 3 more Smart Citations
“…In order to overcome the direction limitation and propose a scalable approach, a more complete end-to-end pipeline for image provenance analysis was described in [52]. That method for graph construction first builds dissimilarity matrices based on local image features, and then employs hierarchical clustering to group nodes and draw edges within the final provenance graph.…”
Section: Related Workmentioning
confidence: 99%
“…As stated in Section 1, relying solely on image content can lead to noisy edge inference. This is especially true for directed edges, which have been shown to be more difficult to derive than undirected edges [11,52]. An option for addressing this is the use of metadata related to the images.…”
Section: Related Workmentioning
confidence: 99%
See 2 more Smart Citations
“…nipulated images. Scouring the internet for "real" tampered images [24] is a laborious process that often also leads to over-fitting in the training process. Alternatively, one could employ a self-supervised process, where detected objects in one image are spliced onto another, with the caveat that such a process often results in training images that are not realistic.…”
Section: Casia Cover Carvalho In-the-wildmentioning
confidence: 99%