Harris算子耦合强度特征的图像复制-粘贴篡改检测算法

目的针对当前较多图像复制-粘贴篡改检测算法主要依靠度量图像的结构特征来实现篡改检测,忽略了图像的强度特征,使其在各种几何变换下难以准确检测出伪造内容,导致检测结果中存在漏检和误检等问题,设计一种基于Harris算子耦合强度特征的图像复制-粘贴篡改检测算法。方法利用Harris算子对图像的特征点进行精确的提取。通过特征点构造圆形特征区域,求取该区域的Zernike矩,通过Zernike矩的大小实现对特征点的描述。随后,利用不同阶数的Zernike矩来描述图像的强度特征和纹理特征,从而构造匹配模型,对图像特征进行粗匹配,并引入RANSAC方法对粗匹配结果进行优化。最后,利用形态学腐蚀与膨胀操作将特征区域进行连通,以确定篡改区域。结果实验结果表明,与已有的图像伪造检测方案相比,所提算法具备更高的检测精度和鲁棒性,在噪声和旋转等变换下仍有更好的检测效果。结论所提技术拥有较高的伪造检测准确性,在图像水印、信息安全领域具有一定的参考价值。

Image Copy-move Forgery Detection Algorithm Based on Harris Operator Coupling Intensity Feature

The work aims to design an image copy-move forgery detection algorithm based on Harris operator coupling intensity to solve the problem that most existing copy-and-paste tamper detection algorithms mainly rely on measuring the structural characteristics of the image to achieve tamper detection, thus ignoring the image strength characteristics, making the algorithm can not adapt to rotation and other tamper detection methods, resulting in the detection results of leakage detection and error detection. The Harris operator was used to extract the feature points accurately. A circular feature region was constructed by feature points, and the Zernike moments of the region were obtained. The feature points were described by the size of the Zernike moments. Zernike moments of different orders were used to describe the intensity and texture features of the image. The matching model was constructed by the size of Zernike moments of different orders to roughly match the image features. The RANSAC method was introduced to further realize the precise matching of image features. The feature area was connected by morphological corrosion and expansion operation to determine the tampering area. From the experimental results, the proposed algorithm had high detection accuracy and roughness and still had better detection results under the noise and rotation, compared with the existing image forgery detection scheme. The proposed technique possesses high forgery detection accuracy and has certain reference value in the fields of image watermarking and information security.