Double JPEG Detection

github: https://github.com/plok5308/DJPEG-torch (2021-02-21 update)

bibtex

@InProceedings{Park_2018_ECCV,

author = {Park, Jinseok and Cho, Donghyeon and Ahn, Wonhyuk and Lee, Heung-Kyu},

title = {Double JPEG Detection in Mixed JPEG Quality Factors using Deep Convolutional Neural Network},

booktitle = {The European Conference on Computer Vision (ECCV)},

month = {September},

year = {2018}

}

Training data

*All JPEG images were compressed by doing quantization of only Y channel using below 1120 quantization tables. We didn't quantize Cb, Cr channel. All images were generated using three raw image datasets. Please refer below paper if you will use my jpeg image dataset.

[34] "D.-T. Dang-Nguyen, C. Pasquini, V. Conotter, G. Boato, RAISE – A Raw Images Dataset for Digital Image Forensics, ACM Multimedia Systems, Portland, Oregon, March 18-20, 2015" .

[35] T. Gloe and R. Böhme, The `Dresden Image Database' for Benchmarking Digital Image Forensics, in Proceedings of the 25th Symposium On Applied Computing (ACM SAC 2010), vol. 2, 1585-1591, 2010.

[36] Bas, Patrick, Tomáš Filler, and Tomáš Pevný. "” Break Our Steganographic System”: The Ins and Outs of Organizing BOSS." International Workshop on Information Hiding. Springer, Berlin, Heidelberg, 2011.


*Extracted 1170 quantization tables were used for Y channel quantization. We didn't use quantization tables were used for Cb,Cr channel. 1 - 100 th are quantization tables of standard quality factors (Q1 - Q100). 101 - 1170 th are quantization tables of non standard quality factors.

Network structure code: https://github.com/plok5308/DJPEG (with tensorflow 1)

Abstract

Double JPEG detection is essential for detecting various image manipulations. This paper proposes a novel deep convolutional neural network for double JPEG detection using statistical histogram features from each block with a vectorized quantization table. In contrast to previous methods, the proposed approach handles mixed JPEG quality factors and is suitable for real-world situations. We collected real-world JPEG images from the image forensic service and generated a new double JPEG dataset with 1120 quantization tables to train the network. The proposed approach was verified experimentally to produce a state-of-the-art performance, successfully detecting various image manipulations.

Network structure

Experiments

Results for local manipulation