Copy-Move Forgery Detection Using Canny Edge Detector and SIFT-Based Blob Analysis

Abstract

This work presents a copy-move forgery detection method based on blobs detection from canny edge-response images, followed by SIFT (Scale Invariant Feature Transform) key points analysis. The SIFT key points are extracted from grayscale images, which are then clustered according to their distance from within generated blobs of that image. Different key points in different blobs are ignored (unique non-duplicated key points) while the same key points in different blobs are matched. The assumption made here is that these duplicate regions are highly indicative of forgery. The proposed method is tested on the MICC-F220, MICC-8Multi, and CoMoFoD public benchmark datasets. The results demonstrate that copy-move forgery detection can be reliably performed on geometrically transformed images, i.e., rotated, scaled, combined and post-processed images such as JPEG compression, blurred, noise added, and contrast adjusted. Our method is also robust on multiple copy-move forgeries.