Research On Image Splicing Tampering Detection Algorithm Based On JPEG Characteristics

In the Internet era,digital images have become one of the main media where people can express and share information.Due to the wide availability of picture editing software,normal individual can manipulate the digital image quickly and easily.Under normal circumstances,people just to pursue beauty.However,in special circumstances,if the criminals maliciously tamper with a picture,which can affect public opinions and even worse a fake picture as an evidence in the court may lead to irreparable damage.Therefore,it is often crucial to ensure that a digital image is authentic and has not been subject to any form of manipulation,and the same as doing research on digital image tampering forensics technology.In this thesis,it focus on the research of splicing detection in the field of digital image forensics.Splicing tampering is a common tampering way,which copies a region of an image,and then pastes it into another image.This thesis presents a comprehensive analysis of the existing algorithms for image splicing tampering detection,and emphatically focus on the iamge splicing tampering localization detection algorithm,also improves two methods:Firstly,facing the problem that some existing methods can not locate the tampering regions,it utilizes a method of image splicing tampering detection based on JPEG compression noise estimation.The method uses JPEG compression to noise images to be detected,extracting the noise feature of splicing tampering regions,and then uses the method of traditional wavelet transform to estimate the noise standard deviation,finally outputs estimation results and shows the tampering region.The experimental results show that the method can accurately detect the tampering region.Secondly,aiming at solving the problem that some detection methods are not accurate to locate tampering region where the image is smooth,it utilizes a method of image splicing tampering detection based on background noise estimation.This method combines Discrete Wavelet Transform(DWT)and skewness characteristics to estimate the noise standard deviation.Firstly it divides the image into blocks,and Wavelet transform is performed on each block and extracts the low-frequency coefficients.Then it estimates the noise standard deviation based on Skewness,and uses method of area elimination to reduce the error,finally uses morphological operations to clearly showthe forgery area.The experimental results show that the method is effective in estimating the noise standard deviation,and has high accuracy in detecting forgery parts especially in local smooth regions.Finally,this thesis summarizes all the work,points out the future research direction,and lists the problem which needs to be solved in this field.