SWT and SIFT based copy-move image forgery detection

Abstract

In our proposed model we have implemented copy-move image forgery detection technique. Copy-move image forgery is one of the types of image forgery where a part of image is copied and then it is pasted in the same image having an intention to make a false image or to hide some important object within the image. Our purpose is to make an efficient and robust solution to this kind of image forgery. Our proposed system consists of few steps: (1) Stationary Wavelet Transform (SWT) is used to decompose the input image into four parts from which approximate image is taken as input for the next step. (2) Scale Invariant Feature Transform (SIFT) algorithm is then run on the approximate image extracted by SWT to extract the key point descriptor features. (3) The descriptor features are clustered using linkage method ward. (4) Clustered key points are compared to take decision whether image is tampered or not. (5) In post processing step false positive removal is done using Random Sample Consensus (RANSAC). Our proposed model after implementations performs 93% accurately over a certain dataset.