Double JPEG Compression Detection by Exploring the Correlations in DCT Domain

TL;DR

This paper presents a novel method for detecting double JPEG compression by analyzing correlations in the DCT domain, using high-pass filtering, PCA, and SVM to improve forgery detection accuracy.

Contribution

It introduces a new algorithm that leverages directional high-pass filtering and machine learning to effectively identify double JPEG compression.

Findings

Effective detection of double JPEG compression across different quality factors

High accuracy achieved with the proposed correlation-based features

Comparable performance to existing methods

Abstract

In the field of digital image processing, JPEG image compression technique has been widely applied. And numerous image processing software suppose this. It is likely for the images undergoing double JPEG compression to be tampered. Therefore, double JPEG compression detection schemes can provide an important clue for image forgery detection. In this paper, we propose an effective algorithm to detect double JPEG compression with different quality factors. Firstly, the quantized DCT coefficients with same frequency are extracted to build the new data matrices. Then, considering the direction effect on the correlation between the adjacent positions in DCT domain, twelve kinds of high-pass filter templates with different directions are executed and the translation probability matrix is calculated for each filtered data. Furthermore, principal component analysis and support vector machine technique are applied to reduce the feature dimension and train a classifier, respectively. Experimental results have demonstrated that the proposed method is effective and has comparable performance.