基于双信息统计与引力聚类的图像篡改检测算法

An Image Forgery Detection Algorithm Based on Dual Information Statistical Coupling Gravitational Clustering

目的为了解决当前图像复制-粘贴篡改检测算法的鲁棒性与检测精准度不佳等问题。方法将图像的颜色信息引入伪造检测过程,提出双信息统计机制耦合引力聚类的图像复制-粘贴篡改检测算法。首先,利用Hessian矩阵来准确提取图像的特征点。然后,利用图像的梯度直方图来描述图像的方向特征,并联合图像的颜色信息,构造双信息统计机制,获取图像的特征向量。计算特征向量间的欧氏距离,构造近似测量模型,对图像特征进行匹配。最后,利用引力聚类方法,实现图像特征点的聚类,精准检测复制-粘贴篡改内容。结果与当前图像复制-粘贴篡改检测方法相比,所提算法具有更高的检测精准度,以及更好的鲁棒性。结论所提方案可以准确检测并定位出伪造内容,在图像水印、信息安全领域具有一定的参考价值。

The paper aims to solve the poor robustness and low detection accuracy of the current image copy-paste forgery detection algorithm. The color information of the image was introduced into the process of forgery detection. An image copy-paste forgery detection algorithm based on dual information statistical mechanism coupling gravitational clustering was proposed. First, the Hessian matrix was used to extract the feature points accurately. Then, the gradient histogram was used to describe the directional features of the image, and the color information of the image was introduced into the feature representation of the image. The double information mechanism was constructed by using the color information and gradient information of the image to obtain the feature vector of the image. An approximate measurement model was constructed by calculating Euclidean distance between feature vectors to match image features. Finally, the clustering algorithm was used to realize the clustering of image feature points and detect the content of copy-paste forgery accurately. The experimental results show that the proposed method had higher detection accuracy and better robustness than the current image copy-paste forgery detection method. The proposed scheme can accurately detect and locate the forged content. It has certain reference value in the field of image watermarking and information security.