特征匹配结合自适应阈值的图像伪造检测

针对现有图像复制移动伪造检测方法对图像中存在同质纹理或均匀区域检测困难、相关参数阈值选择影响检测等问题,提出一种基于特征匹配与自适应阈值的图像复制移动伪造检测方法。首先通过定义和优化成本函数来制定交互作用机制,包括匹配和变换估计步骤;然后在迭代过程中执行步骤,使相关阈值的选择和估计更合理,自动识别和定位伪造区域;最后采用检测系统的相互作用关系,检测并修正步骤中发生的错误。实验结果表明:相比传统的复制移动伪造检测方法,在对伪造图像的像素图像精度水平、旋转和缩放的鲁棒性、检测专业伪造图像以及变换矩阵的精度等不同情况,所提方法效率更高。

An Image Forgery Detection Method Using Feature Matching and Adaptive Threshold

Aiming at the problems existing in image copying movement forgery detection methods,such as the difficulty of detecting homogeneous texture or homogeneous regions,the influence of threshold selection on detection,a novel method for image copying movement forgery based on feature matching and adaptive threshold is proposed.Firstly,the interaction mechanism is formulated by defining and optimizing cost functions,including matching and transformation estimation steps.Then the steps are executed in the iteration process to make the selection and estimation of the relevant threshold more reasonable,identify and locate the forgery area automatically.Finally,the errors in the steps are detected and corrected by using the interaction relationship of the detection system.The experimental results show that the proposed method is more efficient than the traditional detection method of image copying movement forgery in different situations,such as the accuracy level of the pixel image,the robustness of rotation and scaling,the accuracy of detecting professional forged image and the accuracy of transformation matrix.