Region duplication forgery detection in digital images using 2D-DWT and SVD

Abstract

With use of different kinds of software's for high level image processing, nowadays tampering of original digital images is easily done by end users. Therefore this becomes research problem as one cannot know whether present digital image is original or forged. Common type of forgery used is Region Duplication Forgery. In Region Duplication forgery, one part of original image is copied and then pasted over similar image another part. There are many methods presented to locate the Region Duplication forgery detection and localization such as DWT, DCT, SIFT, SVM etc. based method. In literature we have studied some hybrid methods such as DWT-SVD, DWT-PCA etc. However these methods are having limitations regarding to forgery detection accuracy. In this paper, we are presenting new hybrid method Region Duplication Forgery Detection in Digital Images Based on 2D-Discrete Wavelet Transform(2D-DWT) and Singular Value Decomposition(SVD) with goal of improving the performances in terms of hit rate, miss rate and false detection rate. Previously only 2D-DWT based method was presented, however we are further extending this method by using SVD with it so that accurate forgery area can be localized. In proposed method, each layer of RGB image is extracted and then 2D-DWT is performed on forged image. It then divides LL sub-band into overlapping blocks. Perform SVD on each block and sort it into bucket groups to get dominant feature. Performance of forgery detection is measured in terms of hit ratio, miss ratio and FDR.