Image Forgery Detection Using Singular Value Decomposition with Some Attacks

Abstract

To improve the trustworthiness to assess the digital images by identifying authentic images and tampered images, this work is focused on Copy-Move based image Forgery Detection (CMFD) and classification using Improved Relevance Vector Machine (IRVM). In this paper, Biorthogonal Wavelet Transform with Singular Value Decomposition (BWT-SVD)-based feature extraction is applied to find the image forgery. The proposed method begins with dividing the test images into overlapping blocks, and then Biorthogonal Wavelet Transform (BWT) with Singular Value Decomposition (SVD) applies to extract the feature vector from the blocks. After that, the feature vectors are sorts and the duplicate vectors are identified by the similarity between two successive vectors. The occurrences of clone vectors are identified on the basis of Minkowski distance and the threshold value. Then, similarity criteria result in the existence of forgery in images. To classify images into the category of authentic images or forged images, improved version of Relevance Vector Machine (RVM) uses, which leads to efficiency and accuracy of forged image identification process. Performance of proposed scheme tests by performing experiments on CoMoFoD database. The simulation results show that the proposed IRVM scheme attained high performance when compared with existing Copy-Move based image Forgery Detection schemes in MATLAB environment.