A new method for forensics detection based on 2D-histogram and Zernike moments

Pixel mapping transforms such as contrast enhancement and histogram equalization are of most popular methods to enhance the objective properties of image in the forgery applications. In this paper, we introduce a new method for forensics detection. Our data set is a combination of four categories including original images, contrast enhanced images, noisy images, and histogram equalized images. We propose to use Zernike moments of the 2D-Histogram of image in the polar coordinate as features. Moreover, we suggest different features for different classes. We apply two different classifiers namely support vector machine (SVM) and neural network to assign the input image to one of the four mentioned categories. The experimental results demonstrate that the proposed method achieves high classification accuracy.