Forgery Detection by Using Zernike Moments and Nearest Neighbour
With the increasing use of internet, risk for the information is also increasing. High quality software’s are available to make manipulations in the images. These fabrications are so minute that it is not possible to detect them with naked human eyes. It may affect the social, political, economic and personal aspects of society. Wrong prediction of images may also result in jeopardise of judicial system. Therefore, there is a great need to evolve an image authentication system that is capable of detecting forgery in images with high precision rate. Orthogonal image moments are the most popular and most efficient methods to describe an image as they can represent an image with minimum information. They are robust against noise and rotation and can be made invariant to scaling and translation as well. Zernike moments are the most popular orthogonal moment as they are invariant to arbitrary rotation. This paper presents an improved Zernike moment block based technique integrated with generalized nearest neighbour to detect image forgeries. Generalized nearest neighbour is used for feature matching as g2NN is capable of detecting multiple forgeries. Performance of proposed block based Zernike moment method is shown for various attacks like translation, Rotation, scaling and combination of rotation and scaling attacks. Comparison of the proposed technique with the existing techniques is also presented.