Approaches for Digital Image Forgery Detection

A novel framework of Hybrid Neural Networks with Decision Tree (HNN-DT) is introduced in this book, which is efficient for easy training and testing of images for proficient classification of forgery images. Preprocessing by Wiener filter is explained, then the feature extraction process by SURF and PCA to extract the relevant features for classification has been discussed. It then moves to find the matching similarity by Manhattan distance to determine the matching between original and forgery images. In chapter six, the modified Gabor filter and Centre Symmetric Local Binary Pattern (CS-LBP) based feature extraction method is developed to detect the copy-move image forgery based on the texture feature of input images. Hybrid Neural Networks with Decision Tree (HNN-DT) is applied to the feature extraction to classify the forgery images. Four new approaches and extensions to detect copy-move forgery attacks using hybrid feature extraction with efficient classification are presented. All four approaches address the authentic and forgery images classification issue in a non-noisy environment, whereas one out of these also addresses the issue of spliced image forgery detection.