The Research Of Perceptual Hashing And Its Application In The Filed Of Image Authentication

How to identify the image content and ensure its authenticity and integrity has become a hot spot in the field of information security.Perceptual hashing using reasonable means to reduce the dimension of the image,then through the perceptual digest which be able to express the essence of the image content to represent the original image.Perceptual digest has a lot of advantages such as with a small amount of data,robust to perceptual content which is widely used in content identification,content authentication and other fields.After discussing some deficiencies performance of current perceptual hashing algorithms,a new perceptual hashing algorithm is proposed,then lots of research and exploration work were done aiming at applied our perceptual hashing into image authentication field.Innovative achievements in this paper summarized as follows:(1)Aiming at several deficiencies such as poor robustness to rotation,smooth region difficult to extract image features,poor distinction of color tampering of current perceptual hashing algorithms,a novel perceptual hashing algorithm based on WLD features and Zernike moments is proposed.At first,Zernike moments is introduced to normalize the input image for its angle.Then extract WLD features on every image sub-block,after that a series of operations like quantify,coding and encryption were done in order to generate robust perceptual digest.At last,the hash matching problem is measured with Hamming distance.Experiments show that our algorithm has great discrimination and robustness,especially to Gamma correction and rotation our scheme has stronger robust compared with other perceptual hashing algorithms.In addition,our algorithm also can effectively recover or improve some drawbacks performance that current perceptual hashing existed.(2)On the basis of the above work,this paper proposed a new image origin authentication algorithm based on WLD features.First,generate the robustness and security watermark information.Second,divide the input image into fans and embed watermark into DCT-domain of each sub-block.Finally,according to blind watermarking authentication process drawn the certification conclusion.Experimental results show that our algorithm has good robust to test operations,especially to Gamma correction and shear has stronger robust,thus our algorithm suitable forimage source authentication issues.Meanwhile,our scheme is further verified that the WLD perceptual hashing has good robustness.In addition,in authentication process our algorithm has eliminated the dependence on carrier images,more in line with the image origin authentication requirements.(3)In view of the forensic algorithm based on feature points in the smooth area can't extract enough characteristic information,an image content authentication algorithm based on WLD features is proposed.To detect such forgeries,the given image is divided into overlapping sub-blocks of equal size,then represent themselves by WLD orientation component.The feature matrix is sorted using dictionary sort to reducing search space.At last,identified the tamper localization results by main shift vector method and mathematical morphology.Experimental results show that our algorithm can detect one or multiple Copy-Move tampering,and has good robust to Gaussian blur and Gamma correction.Our algorithm showed more stable detection performance in smooth areas than comparison algorithm,we can get the conclusion that it has a certain application value in the field which images are relative smooth.