An Image Forensic Technique Based on SIFT Descriptors and FLANN Based Matching

Abstract

Doctored images are prevalent everywhere since the easy availability of photoediting tools. The research in image forensics focuses mainly on developing techniques that can help discriminate between doctored and legitimate content in an image. There are various kinds of forgeries possible in an image. Here, we present a robust algorithm for copy-move forgery detection. We exploit the simple linear iterative clustering (SLIC) algorithm to divide the source image into non-overlapping, irregular-sized blocks and then use Scale Invariant Feature Transform (SIFT) to determine the feature keypoints with their descriptors. After that, keypoints between blocks are matched using Fast Library for Approximate Nearest Neighbors (FLANN). Forged regions are chalked out accurately employing some morphological operations and analysis using correlation coefficient. To prove the effectiveness of the proposed algorithm, we have tested it on four standard datasets and found out the proposed scheme is performing satisfactorily well. It is helpful after scaling, rotation, and JPEG compression operations too.