相机源识别算法鲁棒性研究  

Research on the Robustness of Camera Source Identification Algorithm

作　　者：张凯 张珣[1] ZHANG Kai;ZHANG Xun(College of Electionics and Information,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学电子信息学院,浙江杭州310018

出　　处：《软件导刊》2021年第9期233-237,共5页Software Guide

摘　　要：相机源识别是数字图像取证技术的重点研究方向之一。为研究相机源识别算法的可靠性与鲁棒性,提取相机噪声信号的频域信息作为分类识别特征,将原始图像切割成不同图像块以破坏图像内部周期性信息,并加上不同密度的噪声掩盖相机传感器噪声信息,结合卷积神经网络进行分类实验。实验结果表明,在不同尺寸下,相机源识别准确率均在95%以上,噪声密度小于0.1时,识别准确率在80%以上。该算法能够适应各种场景,具有良好的可靠性和鲁棒性。Camera source identification is one of the key research directions in digital image forensics technology.In order to study the reliability and robustness of the camera source identification algorithm,the frequency domain information of the camera noise signal is extracted as the basis for classification and identification.In the experiment,the original image was cut into different image blocks to destroy the internal periodic information of the image and was processed by adding noise to weaken the sensor noise information of the camera.Experimental results show that the classification accuracy rate is above 95%in different sizes,and when the noise density is less than 0.1,the classification accuracy rate is above 80%.The results show that the algorithm can be used in various scenarios and be of good reliability and robustness.

关 键 词：相机识别 数字图像处理 傅里叶变换 卷积神经网络 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]