基于核主成分分析的图像模糊篡改检测算法

Blur detection algorithm in image forgery based on kernel principal component analysis

现有的图像模糊篡改检测算法通常提取模糊操作引入的某单一特征进行判断,为更好地提高算法检测效率,提出基于核主成分分析的模糊篡改检测算法。通过奇异值分解提取第一组特征,计算图像二次模糊相关性作为第二组特征,计算图像质量因子作为第三组特征。运用核主成分分析方法实现多特征融合。采用支持向量机进行判断,从而实现模糊篡改检测。实验表明:该算法能够有效地检测数字篡改图像的模糊操作痕迹,并能对模糊篡改区域进行准确定位。

Most existing image blurring forgery detection algorithm consider only one single feature introduced by blurring operation,in order to improve algorithm detection efficiency,propose blur forgery detection algorithm based on kernel principal component analysis（ KPCA）. Through singular value decomposition（ SVD）,extract the first group of features,calculate the secondlary fuzzy correlation as the second group of features,calculate image quality factor as the third group of feature. Multi-feature fusion are achieved using KPCA. Judgement is carried out using support vector machine,so as to realize blur forgery detection. Experimental results show the proposed algorithm can effectively detect blur operation trace of digital tampering image and can accurately locate blur forgery area.