Image Forgery Detection using Multidimensional Spectral Hashing based Polar Cosine Transform

Abstract

The work aims at developing an algorithm for passive detection of copy move forgery in digital images. Accessibility and manipulability of the diverse and sophisticated digital image editing and processing tools has made the integrity of digital images debatable. The sensitivity and importance of the very many applications that revolve around digital images demand absolute authenticity of the digital images. As much simple as image tampering been made by the image editing tools, that much difficult the tampering detection process has become. In this work, a tamper detection method has been proposed that employs polar cosine transform for feature extraction and multi-dimensional spectral hashing for feature matching. The algorithm starts by dividing the image into overlapping patches and then feature vectors are extracted from the patches using Polar Cosine Transform (PCT) and similar patches are identified using multi-dimensional spectral hashing. Finally, post-verification is done to filter out false detection of forged regions. The multidimensional spectral hashing based method uses the outer product Eigen function to improve the performance of similar patch identification. The performance of the algorithm is measured in terms of precision F1 score and recall. Experimental results show that the multi-dimensional spectral hashing based identification of similar patches gives better results compared to some of the existing hashing based techniques. The proposed method proves to be effective in authenticating digital images which are employed in many fields like forensics, medicine, mass media etc.