Image Forgery Detection Using Machine Learning: Exploring Image Quality Assessment and Sharpness estimation

Image forgery techniques pose a significant challenge in the digital era, particularly when it comes to detecting copy-move forgery, which is crucial for ensuring the credibility of digital evidence. This paper introduces an innovative algorithm that uses the image quality assessment metric and sharpness estimation extracted from images to uncover tampering artifacts, such as irregular tampered boundaries and discrepancies in noise patterns between original and manipulated areas. Additionally, the paper utilizes the XGBoost model to effectively distinguish between original and tampered regions, enhancing the accuracy of forgery detection. The proposed approach is evaluated on the MICC-F2000 dataset, consisting of 2000 images, with 1300 original and 700 forged ones. The experimental results, achieving an impressive accuracy of 97.50%, underscore the potential merits of our approach in advancing the domain of image forgery detection.