RRU-Net: The Ringed Residual U-Net for Image Splicing Forgery Detection

Detecting a splicing forgery image and then locating the forgery regions is a challenging task. Some traditional feature extraction methods and convolutional neural network (CNN)-based detection methods have been proposed to finish this task by exploring the differences of image attributes between the un-tampered and tampered regions in a image. However, the performance of the existing detection methods is unsatisfactory. In this paper, we propose a ringed residual U-Net (RRU-Net) for image splicing forgery detection. The proposed RRU-Net is an end-to-end image essence attribute segmentation network, which is independent of human visual system, it can accomplish the forgery detection without any preprocessing and post-processing. The core idea of the proposed RRU-Net is to strengthen the learning way of CNN, which is inspired by the recall and the consolidation mechanism of the human brain and implemented by the propagation and the feedback process of the residual in CNN. The residual propagation recalls the input feature information to solve the gradient degradation problem in the deeper network; the residual feedback consolidates the input feature information to make the differences of image attributes between the un-tampered and tampered regions be more obvious. Experimental results show that the proposed detection method can achieve a promising result compared with the state-of-the-art splicing forgery detection methods.

1. Introduction

Recently, the widespread availability of image editing software makes it extremely easy to edit or even change the digital image content, which is becoming a fearful problem. Struggling to the public trust in photographs, in this paper, our research is specifically focused on the image splicing forgery detection. The splicing forgery copies parts of one image and then pastes into another image to merge a new image as shown in Fig. 1.(a). Because the tampered regions come from other images, the differences of image attributes between the un-tampered and tampered regions exist, such as lighting, shadow, sensor noise, camera reflection and so on, which can be utilized to identify an image suspected of being tampered with and to locate the tampered regions in the forgery image. The existing splicing forgery detection methods have tried to make use of some feature extraction methods for exploring the differences of image attributes. According to the feature extraction methods used in the existing splicing forgery detection methods, they can be mainly classified into two classes: traditional feature extraction-based detection methods and convolutional neural network (CNN)-based detection methods. Figure 1.

An enhanced input features by the residual feedback in the proposed RRU-net. (a) The splicing forgery image; (b) The ground-truth image; (c) The global response of the enhanced input of the first building block in the proposed RRU-net.