Image tampering detection using noise histogram features

With the rapid development of digital image editing tools, the authenticity of digital images becomes questionable in recent years. Image tampering detection is a technology that detects tampered images by using intrinsic image regularities. However, existing intrinsic image regularities are designed for one specific type of tampering operations. When multiple types of tampering operations are used to process a digital image, image tampering detection accuracy is seriously degraded. This paper exploits the re-normalized histogram of noise and noise difference as a new type of intrinsic image regularities, which is computed as the ratio of the histogram to the peak-value. When the image is operated by image tampering, the re-normalized histogram would increase as the peak value of this histogram would decrease. By applying the re-normalized histogram as features, the experimental results show that the proposed method has the ability to detect a wide range of tamper operations, including single type and multiple types of tampering operations without the prior knowledge of tampering operation order, type and parameter.