Detecting Image Forgery Based on Color Phenomenology
Abstract
We propose White Point-Illuminant Consistency (WPIC) algorithm that detects 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 reported in the camera's EXIF file. In manipulated images, the chromaticity coordinates will have a shifted illuminant color relative to the EXIF-reported white point. Absent manipulation, chromaticity coordinates will be in agreement with the specified white point. We detect image manipulations using a convolutional neural net- work operating on a histogram of relevant statistics that indicate the white point shift. We verify this using a real world data set to demonstrate its effectiveness.
Cite
Text
Stanton et al. "Detecting Image Forgery Based on Color Phenomenology." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019.Markdown
[Stanton et al. "Detecting Image Forgery Based on Color Phenomenology." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019.](https://mlanthology.org/cvprw/2019/stanton2019cvprw-detecting/)BibTeX
@inproceedings{stanton2019cvprw-detecting,
title = {{Detecting Image Forgery Based on Color Phenomenology}},
author = {Stanton, Jamie and Hirakawa, Keigo and McCloskey, Scott},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2019},
pages = {138-145},
url = {https://mlanthology.org/cvprw/2019/stanton2019cvprw-detecting/}
}