基于模式噪声的手机图像篡改检测

Cell Phone Camera Image Tampering Detection Based on Pattern Noise

模式噪声作为相机传感器的固有特征,其高频分量与所拍摄的场景无关,并在相机的生命期中相对稳定.因此基于模式噪声的多媒体取证已经被广泛应用在追溯图像来源检测中.然而,由于模式噪声的脆弱性,容易被篡改者从一个相机复制到另一相机上.针对这一问题,首先对基于模式噪声的两种篡改方式进行研究,并通过设定判定阈值进行篡改取证,同时为了解决模式噪声的相关性检测过程中计算量大,算法复杂度高等问题,采用了改进的模式噪声集鉴别方式,大大提高了篡改检测效率.

Sensor pattern noise has been proved to be an inherent fingerprint of digital cameras. Since the main component of it is independent of the environment, it is stable and widely exists in all sensor-based cameras. Therefore, it has been widely used in forgery detection of digital images. However, the sensor pattern noise is vulnerable and easily to be copied from one camera to another. To solve this problem, we first study the two tampering methods which based on sensor pattern noise, then set a determined threshold to realize forgery detection. Meanwhile, in order to settle the problems such as great calculation and high complexity of the algorithm among the correlation detection of sensor pattern noise, this paper introduces a method based on clustered pattern noise and improves the tamper detection efficiency in a large extent.