基于NSCT变换和图像质量评价的拼接图像检测

基于NSCT的多分辨率、局域化、各向异性、平移不变性等特性,文章提出一种新的拼接图像检测方法,利用NSCT和图像质量评价提取出多维图像矩特征以捕获原始图像与伪造图像间的差异,最后利用SVM分类器对图像库进行训练和测试,得到了较好的试验结果。与同类方法相比文中运用了较少的特征向量维数达到了拼接检测目的。

An Effective Method for Blind Detection of Image Splicing with Nonsubsampled Contourlet Transform and Image Quality Evaluation

Nonsubsampled contourlet transform(NSCT) has multi-resolution,location,anisotropy,translation invariance and other characteristics.Sections 1 and 2 of the full paper explain the method of detection mentioned in the title.Their core consists of:(1) we utilize the statistical properties of an image and its quality evaluation to extract its features,thus capturing the differences between original image and fake image;(2) for the statistical properties,we use image blocks to obtain the coefficient matrices with the NSCT;we evaluate the image quality with the block effects.Section 3 uses the support vector machine to train and test the image splicing.The test results,given in Tables 1 and 2,and their analysis show preliminarily that,compared with other detection methods,our detection method can indeed effectively reduce the number of dimensions of the statistical properties and the computational complexity.