Copula-Based Blind Detection of Copy-Move Image Forgery: A Robust Mutual Information Approach

Abstract

Copula functions are powerful statistical tools for modeling the dependency structure between random variables and have been widely applied in domains such as finance, oceanography, and hydrology. However, their application in image processing, particularly for image forgery detection, remains underexplored. This study proposes a novel blind copy-move forgery detection algorithm based on copula-based mutual information, which evaluates statistical dependencies between overlapping image blocks. By leveraging copula theory, the method accurately identifies duplicated regions within a single image without requiring prior knowledge or external references. Experimental results on the CoMoFoD dataset demonstrate that the proposed method achieves high precision, recall, and F1-scores across various manipulation types, including translation, scaling, and rotation, and shows resilience to post-processing operations such as JPEG compression, blurring, noise, and color reduction. Comparative analysis reveals that the copula-based approach outperforms classical methods such as SIFT, SURF, and DWT-SVD. In addition to quantitative performance, qualitative visualizations confirm that the algorithm effectively localizes forged regions in complex scenes with minimal false detections. These findings highlight the potential of copula functions as a robust and efficient framework for digital image forensics.