凸优化耦合传感器模式噪声的图像伪造检测  

Study on Image Forgery Detection Algorithm Based on Convex Optimization Mechanism Coupled Sensor Pattern Noise

作　　者：李景富[1] 张飞[1] 

机构地区：[1]黄淮学院信息工程学院,河南驻马店463000

出　　处：《计算机测量与控制》2015年第5期1678-1681,1685,共5页Computer Measurement &Control

基　　金：河南省重点科技攻关项目研究基金(142102210335);河南省教育厅重点科技攻关项目(13A520786)

摘　　要：现有的图像伪造检测算法主要是借助局部像素与恒虚警率来决策真伪,且忽略了源图像的强烈空间相关性,使算法鲁棒性不佳,难以检测微小尺寸伪造;对此,根据成像传感器的独特随机特性,设计传感器模式噪声检测思想;并提出了凸优化机制耦合传感器模式噪声的图像伪造检测算法;基于光响应非均匀性噪声,联合马尔可夫随机场与贝叶斯规则,设计传感器模式噪声;并构造最佳图像标记像素的先验概率模型;嵌入贝叶斯规则,代替恒虚警率,考虑源图像的强烈空间依赖性,联合整个图像像素,确定最大概率标记像素映射;设计凸优化机制,将图像伪造检测转换为凸问题,提高算法检测效率;并分析了不同伪造区域尺寸对算法检测的影响;仿真结果表明:与当前图像伪造检测算法相比,文章算法具备更好的接收机工作特征;以及更高的检测精度与检测效率。poor robustness and low detection precise were induced by using local pixels to decision authenticity, and ignoring the strong spatial dependence of the source in current image forgery detection algorithms so that can not detect small real area of forgery~ in addition, the high complexity and severe time consuming were existed by the idea of detection step by step adopted in theses algorithms. So the image forgery detection algorithm based on convex optimization mechanism coupled sensor pattern noise was proposed. The best priori probability model of image label pixel was constructed by introducing Markov random field and taking into account the strong spatial dependence of the source ~ embedding the Bayes rule, and jointing entire image pixels to balance the observed statistics and the prior knowledge on the image and penalizing maximum probability label pixel maps for detection the authenticity of the image; taking image forgery detection into convex opti- mization problem, and designed the convex optimization mechanism to improve the detection efficiency. Experimental results show that this algorithm had stronger robust and higher detection accuracy and efficiency.

关 键 词：图像伪造检测 凸优化机制 传感器模式噪声 光响应非均匀性噪声 接收机工作特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]