面向手机图像的来源取证方法  

作　　者：刘橙橙 刁文秀 郭欣 张丽坤 孙建德[1] LIU Cheng-cheng;DIAO Wen-xiu;GUO Xin;ZHANG Li-kun;SUN Jian-de

机构地区：[1]山东师范大学,山东济南250358 [2]不详 [3]北京安瑞中兴科技有限公司,北京100000

出　　处：《信息技术与信息化》2019年第8期12-15,共4页Information Technology and Informatization

基　　金：禁毒情报分析和禁毒网络信息研判技术与装备研究(No.2016YFC0800909)资助

摘　　要：由于信息技术的飞速发展,图像成为了一种重要的信息传递媒体,随之而来的是人们对于图像的篡改和伪造,因此,图像的来源取证成为了近年来的一个研究热点。针对图像来源取证问题,本文首先选取了多种型号手机采集的图像,构建了PHR数据集;然后利用基于多小波变换的特征进行自然图像和生成图像的检测,将自然图像和生成图像进行区分;最后,本文利用了基于小波域特征的手机图像来源判断方法,通过多个SVM分类器的联合应用,实现了对图像的手机来源的辨别。实验结果表明,只要能够选用恰当的特征,就能够高效地区分自然图像和计算机生成的图像,并且能够对手机来源进行判别。Due to the rapid development of information technology,images have become an important information transmission medium,which is followed falsification and forgery of images.Therefore,the source evidence of images has become a research hotspot in recent years.In order to solve the problem of image source forensics,this paper first selects the images collected by various models of mobile phones,and constructs the PHR dataset.Then,the natural image and the generated image are detected by using the features based on multiwavelet transform to distinguish the natural image from the generated image.Finally,this paper makes use of the wavelet image feature based mobile phone image source judgment method,through the joint application of multiple SVM classifiers,realizes the identification of the mobile phone source of the image.The experimental results show that as long as the appropriate features can be selected,natural images and computer-generated images can be efficiently distinguished,and the source of the mobile phone can be discriminated.

关 键 词：图像来源取证 多小波变换 小波域特征 SVM分类器 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]