Detection and Localization of Image Forgeries using Resampling Features\n and Deep Learning

Resampling is an important signature of manipulated images. In this paper, we\npropose two methods to detect and localize image manipulations based on a\ncombination of resampling features and deep learning. In the first method, the\nRadon transform of resampling features are computed on overlapping image\npatches. Deep learning classifiers and a Gaussian conditional random field\nmodel are then used to create a heatmap. Tampered regions are located using a\nRandom Walker segmentation method. In the second method, resampling features\ncomputed on overlapping image patches are passed through a Long short-term\nmemory (LSTM) based network for classification and localization. We compare the\nperformance of detection/localization of both these methods. Our experimental\nresults show that both techniques are effective in detecting and localizing\ndigital image forgeries.\n