SIFT BASED IMAGE TAMPERING DETECTION USING OPTIMIZATION

Abstract

Nowadays digital pictures are an effective and broadly used communication medium. They importantly affect our life. Because of the widely used editing softwares like Photoshop, it has turned out to be generally simple to make fake pictures. So there is a need to develop strategies that will detect all kinds of tampering in the images. In this thesis we propose a technique to consequently recognize the copied regions in advanced pictures. The nearness of copied regions in a picture may imply a type of fabrication called copy–move forgery. The proposed technique coordinates both keypoint based and block based phony detection strategies. To start with, the segmentation process fragments the host picture into non-overlapping and irregular squares utilizing SLIC Algorithm. At that point, the feature points are separated from each part as block elements, and the block elements are matched to find the labelled highlight points; this technique can roughly show the presumed tampering regions. To distinguish the forged areas all the more precisely, we propose the tampered region extrication method, where the feature points are replaced with little superpixels as highlighted areas and afterward combines the neighbouring blocks that have comparable nearby shading highlights in the feature blocks to produce the merged areas. At last, it implements the morphological operation to the merged locales to obtain the identified tampered regions. However, for some Copy Move Forgery (CMF) pictures, these methodologies can't create acceptable detection results. For example, the quantity of the coordinated key-points might be too less to end up being a CMF picture or to produce an exact outcome. Once in a while these methodologies may even deliver errors. To take care of the issue, a novel approach named as CMF Detection with Cuckoo Search Optimization (CMFD-CS) is proposed in this thesis. CMFD-CS coordinates the Cuckoo Search Optimization (CS) algorithm into the SIFT-based structure. Experimental results demonstrate that CMFD-CS has better execution.