基于统计特征和马尔可夫特征的图像拼接盲检测

Method for Image Splicing Detection Based on Statistical Features and Markov Feature

数字图像真实性检测在司法鉴定等领域有着重要的作用。常见的图像拼接篡改会降低图像像素直接的相关性,这可以通过一些统计特征反映出来。采用特征提取-分类的方法,提取矩特征、基于二维相位一致性的统计特征,结合DCT域的马尔可夫特征,利用SVM分类器进行分类,实现了拼接图像的盲检测。实验结果表明,该方法有较好的鉴别准确率,可达91.75%。

The authenticity detection of digital images is of great significance in judicial identification. Image splicing is used very often as a falsification method and it will certainly decrease the correlation between image pixels which can he reflected by some statistical features. Combined with Markov features,a new approach which uses moment and phase congruency to extract features was proposed and SVM classifier was used to judge the category an image belongs to. The experimental results show that the proposed method can achieve a detection accuracy of 91.75 %.