Computational data analysis for first quantization estimation on JPEG double compressed images

In Proceedings of the International Conference on Pattern Recognition

Sebastiano Battiato^1,2, Oliver Giudice^2,3, Francesco Guarnera^1,2, Giovanni Puglisi⁴

¹ Department of Mathematics and Computer Science, University of Catania, Viale Andrea Doria 6, Catania 95125, Italy
² iCTLab s.r.l. Spinoff of University of Catania, Italy
³ Banca d'Italia, Rome, Italy
⁴ Department of Mathematics and Computer Science, University of Cagliari, Via Ospedale 72, Cagliari 09124, Italy

battiato@dmi.unict.it, oliver.giudice@bancaditalia.it, francesco.guarnera@unict.it, puglisi@unica.it

DOWNLOAD PAPER

@article{battiato2021computational,
title={Computational data analysis for first quantization estimation on JPEG double compressed images},
author={Battiato, Sebastiano and Giudice, Oliver and Guarnera, Francesco and Puglisi, Giovanni},
booktitle={2020 25th International Conference on Pattern Recognition (ICPR)},
pages={5951--5958}
year={2021}
organization = {IEEE}
}

JPEG First Quantization Estimation

This program is free software: it is distributed WITHOUT ANY WARRANTY.

If you are using this software, please cite:

Battiato, Sebastiano and Giudice, Oliver and Guarnera, Francesco and Puglisi, Giovanni: “Computational data analysis for first quantization estimation on JPEG double compressed images", In Proceedings of the International Conference on Pattern Recognition, 2021

The software estimates the first 15 quantization factors (in zig-zag order) of an aligned Double compressed JPEG image.

Given the double JPEG compressed image I, the software read the second quantization matrix Q2 from I and it simulate all possible double compressions employing a constant matrix for each value between 1 and max (set inside the file v1.py) for the first compression and Q2 for the second one.

128_128.jpg: 128X128 image compressed with QF1=60 e QF2=90

4288_2848.jpg: 4288X2484 image compressed with QF1=60 e QF2=90

The JPEG files (128_128.jpg and 4288_2848.jpg) have the first 15 quantization factors (in zig-zag order) equal to [13, 9, 10, 11, 10, 8, 13, 11, 10, 11, 14, 14, 13, 15, 19]

v1.py: python file to predict the quantization factor (function get_coefficients_first_compression in the main).

jpeg: executable file for DCT values extraction from file (without IDCT); it needs the executable permissions.

We tested our codes on Python 2.7 and Python 3 under Ubuntu 16.04 and 18.04 (64 bit).

Try the code

Just launch the installation of requirements.txt and then execute:

python simulate_and_predict.py -i image_name.jpg