基于DWT和聚类方法的图像复制粘贴伪造检测方法研究

Research of Image Copy-Move Forgery Detection Based on DWT and Cluster Method

随着图像处理软件的发展,修改图像变得越来越容易,其中常见的修改方式是图像内部的复制粘贴修改。如何快速、有效检测图像的复制粘贴修改区域,已经成为一个急需解决的问题。目前,很多检测技术虽然可以检测出复制粘贴区域,但是其运算量较大,检测速度较慢。基于此,提出了一种基于离散小波变换和聚类方法的数字图像区域复制粘贴检测方法。该方法首先对图像进行离散小波变换,降低图像的数据量;其次,分别对三个子图进行聚类;最后,综合四个子图的聚类结果,得到图像中的复制粘贴伪造区域。实验结果表明,该方法能快速、有效检测出复制粘贴修改区域,并且具有较高的准确率。

With the development of image processing software,it is more and more easy to modify the image.The common modification method is copy-paste modification inside the image.How to quickly and effectively detect the copy-paste modification area of an image has become an urgent problem to be solved.At present,although many detection techniques can detect the copy and paste area,but its computational complexity is large and detection speed is slow.Based on this,a region copy and paste detection method of digital image based on discrete wavelet transform and clustering method is proposed.Firstly,the image is decomposed by discrete wavelet transform to reduce the amount of data;secondly,three sub-graphs are clustered separately;finally,four sub-graphs are synthesized to obtain the duplicate paste forgery region of the image.The experimental results show that this method can detect the copy-paste modification area quickly and effectively,and has a high accuracy.