Detecting Image Forgery with Color Phenomenology

Date of Award

2019

Degree Name

M.S. in Electrical Engineering

Department

Department of Electrical and Computer Engineering

Advisor/Chair

Advisor: Keigo Hirakawa

Abstract

We propose a method that is designed to detect manipulations in images based on the phenomenology of color. Segmented regions of the image are converted to chromaticity coordinates and compared to the white point, D65. If an image had been manipulated, the chromaticity coordinates will have a shifted white point relative to D65, the accepted average white point. We classify the image forgery using a convolutional neural network using a histogram of relevant statistics that indicate the white point shift. We verify this using a real world data set to demonstrate its effectiveness.

Keywords

Electrical Engineering, Engineering, Image Forensics, Deep Learning, Convolutional Neural Network, CNN, Chromaticity, Multimedia Forensics

Rights Statement

Copyright © 2019, author

Link to Full Text

Share

COinS