Distinguishing Computer-generated Graphics from Natural Images Based on Sensor Pattern Noise and Deep Learning

Computer-generated graphics are images generated by computer software. The rapid development of computer graphics technologies has made it easier to generate a photorealistic computer graphics, and these graphics are quite difficult to distinguish from natural images by our naked eyes. In this paper, we propose a method based on sensor pattern noise and deep learning to distinguish computer-generated graphics (CG) from natural images (NI). Before being fed into our convolutional neural network (CNN)-based model, these images—including the CG and NI—are clipped into image patches. Furthermore, several high-pass filters (HPF) are used to remove low-frequency signal, which represents the image content. These filters are also used to enhance the residual signal as well as sensor pattern noise introduced by the digital camera device. Different from the traditional methods of distinguishing CG from NI, the proposed method utilizes a five-layer CNN to classify the input image patches. Based on the classification results of the image patches, we deploy a majority vote scheme to obtain the classification results for the full-size images. The experiments have demonstrated that: 1) the proposed method with three high-pass filters can achieve better results than that with only one high-pass filter or no high-pass filter. 2) the proposed method with three high-pass filters achieves 100% accuracy, although the natural images undergo a JPEG compression with a quality factor of 75.