项目名称: 基于篡改评价模型的图像真伪盲鉴别方法研究
项目编号: No.61305046
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 陈海鹏
作者单位: 吉林大学
项目金额: 25万元
中文摘要: 图像真伪问题已渗透到政治、经济、军事、文化等各大领域,严重影响了人们正常的生活和生产秩序。目前数字图像盲鉴别算法需已知篡改手段,仍处于零散技术的研究阶段。基于此,本项目深入研究基于篡改评价模型的图像盲鉴别方法及应用。建立基于自然图像特征的篡改评价模型,定性分析篡改手段。针对重获图像,研究基于成像过程模型和物理特征模型的盲鉴别算法;针对拼接篡改图像,研究基于自然图像特征一致性模型的盲鉴别算法;针对复制粘贴篡改图像,构造SIFT关键点标记图特征向量,研究具有旋转、缩放不变性的盲鉴别算法。将图像篡改评价模型与盲鉴别算法相结合,在定性分析篡改手段的基础上,精确定位篡改区域。 本项目的预期技术指标为:图像篡改手段定性判断的准确率达80%,盲鉴别算法的平均正检率达85%。本项目不仅为数字图像盲鉴别技术的研究开辟了新的视角,引入了新的理论框架,而且提高了数字图像盲鉴别技术在司法取证等领域的实用性
中文关键词: 图像篡改;盲鉴别;篡改评价模型;拼接盲鉴别;复制粘贴盲鉴别
英文摘要: The issues of image authenticity have infiltrated politics, economy, civilization, military affairs and so on, which results in a lot of negative effects. Nowadays, accurate blind identifications of digital images depend on the knowledge of the used tampering method. However there is little work on developing model for evaluating which tampering method has been used on certain images. Thus this project seeks to develop (i) an evaluation model for analyzing which tampering method is used on the image and (ii) identification algorithms for identifying which part of the image is being tampered. Evaluation model of image tampering will be established to analyze the tampering operations qualitatively. As for identification algorithms, blind identification algorithm will be researched and realized for recaptured images, which is based on the model of imaging process and the model of physical characteristics. For spliced images, we will research on the blind identification algorithm based on the model of the consistence of natural image features. For copy-paste images, the feature vectors of SIFT key points will be constructed, which are made up of the marked graph, and then the blind identification algorithm robust to rotation and scale will be researched. Finally, we will combine the evaluation model of image tamp
英文关键词: Image tampering;Blind forensic;Tampering evaluation model;Blind detection of image splicing;Blind detection of copy-move image