一种用于图片来源检测的相机参考模式噪声提取算法

摘要

在传统相机参考模式噪声提取算法中,采用多张图片传感器模式噪声的平均值作为相机的参考模式噪声,但其未滤除插值颜色产生的噪声且未考虑不同类型的图片在计算相机参考模式噪声时所占比重,导致提取的相机参考模式噪声效果较差.为此,根据彩色滤波矩阵噪声方差的特点以及图片纹理和亮度的差异,提出一种改进的相机参考模式噪声提取算法.通过滤除插值颜色产生的噪声获取所有图片的真实传感器模式噪声,按照图片纹理和亮度乘积得到每张图片传感器模式噪声在相机参考模式噪声中所占比重,从而计算出相机的参考模式噪声.实验结果表明,与传统相机参考模式噪声提取算法相比,该算法具有更高的图片来源检测准确率.%Camera reference pattern noise is computed by averaging sensor pattern noise values of all images in the traditional camera reference pattern noise extraction algorithm.However,the traditional method cannot filter the interpolation noise and not consider the proportion of different types of images.To improve the extraction effect of the camera reference pattern noise,this paper proposes an improved camera reference pattern extraction algorithm based on the characteristics of noise variance and texture and lightness differences among images.Firstly,it extracts all image sensor pattern noise and removes interpolation point to extract real image sensor pattern noise.Then it uses the product of image texture and lightness as the weight of each image sensor pattern noise in the camera reference pattern noise.Finally,it uses a weighted average to estimate camera reference pattern noise.The experimental results show that this algorithm achieves higher accuracy in image source detection compared with traditional camera reference pattern noise extraction algorithm.