Automatic Detection of Image Morphing by Topology-based Analysis

Jassim, Sabah A. and Assad, A. (2018) Automatic Detection of Image Morphing by Topology-based Analysis. In: 26th European Signal Processing Conference (EUSIPCO), 3-7 September 2018, Rome, Italy.

[img] Text
Automatic Morph detection by Topology based methods_June2018-Accepted at EUSIPCO 2018 (002).pdf
Restricted to Registered users only

Download (527kB) | Request a copy

Abstract

Topological Data Analysis (TDA) is an emerging framework for the understanding of Bigdata. This paper investigates and develops a TDA approach to image forensics that exploits the sensitivity to image tampering of a variety of persistent homological invariants of simplicial complexes constructed for certain automatically computed image texture landmarks. For each image, we construct sequences of simplicial complexes, whose vertices are the selected set of landmarks, for a sequence of distance thresholds and use a variety of homological invariants (e.g. number of connected components) to distinguish natural face images from morphed ones. We shall demonstrate the richness of TDA in dealing with image tampering by testing the performance of this approach on a large benchmark image dataset of passport photos in detecting various known morphing attacks.

Item Type: Conference or Workshop Item (Paper)
Additional Information: DOI 10.23919/EUSIPCO.2018.8553317
Uncontrolled Keywords: Image Morphing attacks; TDA; Simplicial Complexes; Local Binary Pattern; Persistent Homology
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: School of Science > School of Computing
Depositing User: Rachel Pollard
Date Deposited: 23 Aug 2019 13:45
Last Modified: 23 Aug 2019 13:45
URI: http://bear.buckingham.ac.uk/id/eprint/405

Actions (login required)

View Item View Item