The Research Of Passive Forensics Against Image Seam Carving

The prevalence of image acquisition devices and image processing software have given rise to large amounts of doctored images with no obvious traces.If these images are to be used by our judicial system,the news media or the field of science,it will bring immeasurable negative impact on social order et.al.Thus,there is a great demand for automatic forgery detection algorithms that can identify the trustworthiness of a candidate image.Passive image forensics does not require any auxiliary data such as watermarks or signatures and directly identifies the authority of images based on inhere statistical properties.Especially,seam carving used to be a normal mean to adapt image display on diverse terminals.However,it can also be a useful tool for object removal.Therefore,passive forensics against seam carving confronts both technical challenges and great application potential.This paper focused on forgery detection against seam carving,and the innovations and contributions are as follows:In order to improve detection precision and enhance robustness to post-processing,this paper presents an efficient forgery detection algorithm for seam carving.Since seam carving changes the local texture in an image,local texture descriptor,i.e.,local binary pattern(LBP)is introduced to reveal the local texture artifacts when seams are removed from an image.Twenty-four statistical features are extracted from the LBP image to measure the energy bias and noise difference.Among them,six energy features are newly defined in terms of half-seam,instead of the whole seam.Finally,support vector machine(SVM)classifier is exploited to determine whether an image is original or suffered from seam carving.Experimental results show that the proposed approach achieves desirable detection accuracy,which improves about 5%on average compared with the state-of-the-art approaches.To enhance robustness to seam carving with low scaling ratio,this paper presents a blind detection approach for seam carved image with low scaling ratio(LSR).We observe that when few seams are deleted from an original image,its spatial and spectral entropy(SFE)distribution is significantly changed.Forty-two features are designed to unveil the statistical properties of SFE in terms of centralized tendency,dispersion tendency and distribution tendency.They are combined with the improved LBP-based energy features(54 features)to form ninety-six features.Finally,SVM is exploited as pattern classifier to determine whether an image is original or has been suffered from Seam carving.Experimental results show that the proposed approach achieves superior detection accuracy over state-of-art works,especially for those resized image by Seam carving with low scaling ratios.Moreover,it is robust against JPEG compression and Seam insertion.Up to present,passive forensics against Seam carving is still in its preliminary exploration stage.It is hoped that the present study maybe promote the development of passive image forensics.