基于旋转灰度特征与位移约束的图像伪造检测算法

Image Forgery Detection Algorithm Based on Rotation Gray Feature and Displacement Constraint

当前图像伪造检测方法主要依靠近邻比值法来实现图像内容的真伪决策,由于近邻比值法的判断条件单一,该类算法的检测精度不佳.因此,本文提出了旋转灰度特征耦合位移约束的图像伪造检测算法.首先,根据图像像素点,构造十字约束法则,对FAST算法予以改进,以提取待检测图像中的特征点;然后,利用像素点的梯度特征形成直方图,通过直方图峰值获取特征点的主方向;再用特征点的灰度值来构造旋转灰度特征模型,用于获取特征向量,生成特征描述子;用特征点的位置以及角度特征,构造了位移约束规则,并且在位移约束规则下,通过归一化互相关函数对特征点的相似性进行度量,完成特征点匹配.最后,引入均值漂移模型,对图像中的伪造内容完成区域定位,实现图像的伪造检测.实验结果表明:与当前图像伪造检测算法相比,本文算法不仅具有更高的检测精度以及检测效率,而且还具有更好地鲁棒性能.

In view of the current image forgery detection method to achieve image matching and image forgery detection mainly relies on the nearest neighbor ratio method, when the detected image exists in the complex forge content image, this method appears more error detection point and detection of defects such as time-consuming. So an effective image forgery detection algorithm based on rotation gray feature and displacement constraint was proposed in this paper. First of all, the FAST algorithm is used to construct the cross constraint rule, and then the improved algorithm is used to extract the feature points in the image to improve the detection efficiency. Then, the histogram of the pixel is used to form the histogram, and then the main direction of the feature point is obtained by the histogram peak. Then, using the gray value of the feature points to construct the rotation gray feature model, which is used to obtain the feature vector and generate the feature descriptor. Finally, the position of the feature points and feature point, the formation of displacement constraint rules, and displacement constraint rules using normalized cross-correlation function to measure the similarity of feature points, feature point matching. The mean shift model is used to locate the forgery in the image, and the forgery detection is completed. The experimental results show that the forged images were detected when compared with the current image forgery detection algorithm; this algorithm has higher detection precision and efficiency, but also has better robust performance.