Research On The Methods Of Image Filtering Identification And Resamping Detection

The rapid development of digital image processing technology and the widespread use of graphics/image processing softwares(such as Photoshop,Meitu,and so on)are the "double-edged sword".Not only it brings convenience to people,also it makes very easily to edit,modify,even fake the image content.As a result,many events involving images content tampering occur constantly in recent years.Therefore,there is urgent requirement to find practical ways to ensure the authenticity and integrity of image content.Digital image forensics is exactly such technology.Recently,digital image forensics has become an important research topic in the field of digital media content security protection.Considering the image filtering and image re-sampling are not only common image forgery technologies,and typical anti-forensics approaches,we study the passive forensics methods for digital image filtering detection and image re-sampling detection in this paper.The major works are as follows:We put forward a digital image filtering detection method based on Local Binary Pattern(LBP).According to the differences of image LBP feature pattern before and after filtering operation,we explore the changing pattern of image LBP feature values and find statistical discipline then we define the discrimination rule by analyzing the number relationship of special LBP binary codes in digital image and establish two filter operation detection algorithms: Algorithm I is a discriminating algorithm that can distinguish a median(average,Gaussian)filtered image from a pair of given original image and its median(average,Gaussian)filtered version.Algorithm II is a recognition algorithm that can identify whether or not a given image has undergone median filtering.Compared with the Fourier spectrum of original image,the Fourier spectrum of re-sampled image will present obvious changing.Based on this fact,we propose an image re-sampling detection method based on Fourier transform.In the proposed method,we carry ondiscrete Fourier transform firstly to grayscale images,then observe the differences of Fourier transform spectrum and calculate the standard deviation.Finally,we define the threshold value according to the specificity of the standard deviation caused by the different sampling factors.Using this threshold we can detect image up-sampling and down-sampling operation respectively,and determine whether or not the image has undergone re-sampling operation.Different from traditional image filtering detection methods and image re-sampling detection methods that are based on statistical learning and classification,the proposed methods do not need large-scale training sample database and any classifiers.The experimental results show that the detection rates of our filtering detection methods are above 98.13% and 99.24%,respectively.And the detection rate of our resampling detection method is above 84.9%.Compared with the state-of-the-art algorithms,the proposed methods have obvious advantages.