项目名称: 基于模式识别技术的数字图像盲取证研究
项目编号: No.60803136
项目类型: 青年科学基金项目
立项/批准年度: 2009
项目学科: 金属学与金属工艺
项目作者: 卢伟
作者单位: 中山大学
项目金额: 19万元
中文摘要: 近些年数字图像盲取证已经成为国际上研究的前沿领域,它为分析数字图像的真实性和来源提供了一种新颖和有效的途径。本项目以检测数字图像伪造为背景,利用模式识别技术,主要研究基于数字图像区域复制篡改和拼接的取证理论,从模型和技术上研究揭示数字伪造的方法。通过三年的深入研究,提出了几种新颖独特的能够揭示数字图像伪造的理论和算法,包括基于Harris Affine不变性检测和匹配、基于DCT和DWT系数的马尔科夫转移概率矩阵特征分类、基于模糊度量特征的分类、基于高阶局部自相关统计特征的分类等方法。研究结果在目前主流学术界具有较好的评价,性能达到并部分超过了目前已知的最好方法。此外,通过对前人的一些方法进行借鉴和改进,并综合我们提出的方法,在算法设计和理论分析上进行了优化,性能和效率都获得了大大地提高。同时,设计了一个数字图像取证平台,集成了目前主流的一些数字图像取证算法,建立了一套独立的数字图像篡改检测数据库,获得了主流学术研究界的认可。通过研究,在数字图像盲取证领域取了较好的成绩,如果能获得NSFC新的资助,相信对未来数字媒体取证具有重要的意义。
中文关键词: 数字图像取证;篡改检测;区域复制;图像拼接;模式识别
英文摘要: Image tampering is very common and fundamental nowadays. To recover people's trust in digital images, the detection of image tampering is in great need. This program aims at detecting both region duplication tampering and image splicing blindly by using pattern recognition techniques. Some novel and effective methods has been developed by us, including Markov based approach in DCT and DWT domain, Harris-Affine region detection and matching, classification by blurring measurement, and higher local autocorrelation statistics etc. Some existing methods are also improved and optimized by combing with our proposed methods. The experiment results demonstrate that our proposed approaches can outperform the state of the art when applied to some popular image tampering detection evaluation dataset. At the same time, we developed a digital image forensics platform that integrating some main digital image forensic algorithms, and build a new digital image tampering detection dataset, which are all approbatory by the academic circles. Good outcome has been made by this program, we hope further funding by NSFC, and achieve more excellent contributions.
英文关键词: Digital image forensics; Tampering detection; Region duplication; Image splicing; Pattern recognition