| With the rapid development of smart devices,everyone can become a disseminator and receiver of information,with the ability to shoot,edit and disseminate information.The tampering of information will lead to a series of problems,and the authenticity of information can only be guaranteed through authenticity detection of images,making digital forensics indispensable.In this paper,we analyze and research based on image authenticity detection,focusing on two tampering methods,copy-paste and splicing of images.The main works are:(1)The Y-SIFT copy-paste detection algorithm combined with edge detection is proposed.The SIFT-based image copy-paste detection algorithm only detects and locates feature points in grayscale space,ignoring the important information in color space,thus leading to problems such as inaccurate localization.To address this problem,a Y-SIFT copy-paste detection algorithm combined with edge detection is proposed.By fusing YCr Cb color information and using color feature descriptors for feature extraction of images,the pixel differences on the edges of tampered regions are analyzed at the same time,and two groupings are innovatively proposed to be performed simultaneously.Experiments prove that the method improves the detection accuracy and localization accuracy of tampered images.(2)An image stitching detection algorithm combining LBP and PCA is proposed.After the post-processing operation of the stitched and tampered image,the statistical attributes of the tampered region are less different from those of the original image,which leads to the low accuracy of a single feature extraction method for detecting this type of tampering.The specific implementation steps of the image stitching detection algorithm combining LBP and PCA proposed in this thesis are as follows: firstly,superpixel segmentation is performed on the image,then multiple feature information of image tampering is extracted by extracting LBP texture features of the image and using PCA to estimate the noise level for detection,followed by classification of the image according to the inconsistency of texture information and noise level of the tampered image,and finally,using morphology to achieve region localization of the tampering operation.The experimental results show that the method has good results in terms of the accuracy of localization. |
