Research On Color Image Copy-move Forgery Detection Based On Superpixel And Quaternion

With the rapid development of Internet information technology,the number of digital images has increased significantly.At the same time,the processing of digital images has become more and more simple,resulting in a large number of fake images in various fields.Therefore,the authenticity of images is seriously questioned.Copy-move forgery is a relatively common method of fakery,so Copy-Move Forgery Detection came into being.In recent years,many researchers have proposed a large number methods of CopyMove Forgery Detection,which can accurately detect tampered areas to a certain extent.However,most of these methods are aimed at grayscale images.For the processing of color images,the color information of color images and the correlation between color channels are ignored,and the image content can not be completely expressed.In addition,these methods usually divide the images into overlapping regular blocks,but a large number of overlapping blocks lead to low efficiency of feature extraction and matching,and the regular blocks are weak to geometric transformation.In view of the above problems,this paper has mainly studied the following two aspects after summarizing the research status:(1)A Copy-Move Forgery Detection algorithm based on shape features of superpixel is proposed.Firstly,entropy rate superpixel segmentation method is used to segment images and the stable feature points are extracted.Then a novel shape coding scheme is proposed to extract superpixel shape features,which are merged with the feature points to estimate the suspected forged regions.Finally,the suspicious forged regions are segmented into superpixels again to be matched to accurately locate tampered areas.Experimental results show that the proposed method has the ability to resist geometric transformation,noise,blur and JPEG compression.(2)An algorithm of Copy-Move Forgery Detection for color images based on superpixel and quaternion is proposed.Firstly,a method for adaptively dividing superpixels based on wavelet contrast is proposed.Then,quaternions are used to represent the color image,and the stable low-order quaternion Exponent moments with color invariance and geometric invariance are extracted as the feature vectors of superpixels fused with shape coding to estimate the suspicious tampered regions;finally,the suspicious tampered regions are divided into circular blocks and the quaternion Exponent moment features of the circular blocks are extracted,and the tampered regions are accurately located by matching the features of circular blocks.Experimental results show that the proposed method shows good performance in resisting geometric transformation,JPEG compression,noise,etc.,and effectively improves the detection efficiency.