Forgery detection in medical images with distinguished recognition of original and tampered regions using density-based clustering technique

Abstract

Telemedicine is one of the most significant part of the medical field. It enables transfer of medical images over the internet which offers telediagnosis facility. Due to transfer of medical images from different channels, detection of their authenticity is a crucial research concern. Copy-move forgery is one of the most popular technique for image manipulation. We propose a method for copy-move forgery detection in medical images. Our method apply boundary extraction followed by Laplacian blob detection from the image to identify regions with similar properties. Further, we utilize Good Features To Track (GFTT) and BinBoost techniques for keypoint extraction and descriptor computation, respectively. Similar descriptors are identified using Hamming distance-based nearest neighbor search. Clustering over keypoints is performed using Ant Colony Density-based Clustering (ACDC) technique. We employ Fast Sample Consensus (FSC) technique for selection of correct matches and removal of imprecise keypoints. Further, we use correlation map generation for localization of manipulated regions. Experimental outcomes display that our technique can efficaciously identify tampered medical images even when manipulated images are sustaining various geometrical and post-processing attacks. Proposed technique can also efficaciously distinguish between original and tampered region within forged medical images using Haralick texture features.