网函数插值法在数字图像盲检测中的应用

Application of Nets Function Interpolation to Blind Detection of Digital Forgeries

针对计算机生成图像和拼接图像伪造,提出一种基于网函数插值法的数字图像盲检测算法.该算法首先利用Coons型分形曲面片对待检测图像进行预测.然后计算待检测图像和预测图像之间的差异,并提取待检测图像、预测图像及其差值的统计特征.最后,根据这些特征利用支持向量机判定待检测图像是否为自然图像.实验结果表明:与基于高阶统计特征和几何不变量的算法相比,该算法具有更高的效率、准确率和稳定性.

Targeting computer graphics and spliced images,this paper proposes an image blind detection algorithm based on nets function interpolation.Firstly,we get a predicted image by predicting the detected image with Coons fractal patches.Secondly,we calculate the difference between the predicted image and the detected image,and extract the statistical features from both images and their difference.Finally,we determine whether the detected image is original by support vector machine with the statistical features.The experimental results show the proposed algorithm performs better than the algorithms based on higher order statistical feature and geometry invariants in term of efficiency,accuracy and stability.