现存的大部分篡改检测方法对篡改区域的几何变化检测比较敏感，针对该问题，提出一种利用特征图像块精细化自动检测篡改区域的数字图像取证方法，该方法适用于反射、旋转、缩放区域和JPEG压缩定位。首先将重复区域的像素映射到对数极坐标上。然后沿轴，利用反射和旋转产生一维不变描述符。此外，运用每个单独块中提取的特征向量来减少每个阶段的计算时间。最后利用一个精细化阶段复制几何变换后的重复区域。实验对尺寸为24×24和32×32的块进行检测，比较两种情况下获得的定位结果可知，导致较高的真阳性率的测试同时也会导致较低的假阳性。此外，对篡改和未篡改的图像分别进行检测实验，结果表明，与其他算法相比，该算法对几何变换后的图像具有较高的篡改定位准确率和较低的错误匹配率。%Most of the existing tampering detection methods are sensitive to the geometric transformation detection of the tampering area,for this problem,a digital image forensics method of using feature image blocks to detect the tampering area re⁃finedly and automatically is proposed. This method is suitable for area reflection,rotation and scaling,and JPEG compression and positioning. The pixels of the repeat area are mapped to the log⁃polar,and then along the axis,the invariant one⁃dimensional descriptors are generated by reflection and rotation. Each single block is used to extract the feature vectors to reduce the compu⁃tation time of each stage. After that the repeat area after geometric transformation is copied with a refined stage. The blocks whose size are 24×24 and 32×32 are detected with experiments. The positioning results obtained from the two situation show that the test can lead to high true positive rate while causing low false positive rate. The tampered and non⁃tampered images were de⁃tected with experiments. The experimental results show that,in comparison with other algorithms,the proposed algorithm has high tampering positioning accuracy and low false match rate of the image after geometric transformation.
展开▼
