Research On Passive Forensics Methods For Image Splicing/composite Detection

With the rapid development of digital technology,digital image editing software and tools are everywhere,which makes malicious tampering of images content becomes more easy and convenient.The authenticity and integrity of image content are seriously threatened.Image splicing/composing is a common method of image manipulation.The faked images generated by splicing/composing exist in various fields such as political,economic,military,law,news,science and technology,and so on,and causing serious adverse effects.In this paper,we mainly study the technique of image splicing/composing detection that is based on passive forensics,and the main works are as follows:An image splicing/composing detection method based on Color Filter Array interpolation pattern is proposed.The basic idea is as follows: for the image that is taken by digital camera with CFA model,the pixel correlation model introduced by CFA interpolation is consistent and continuous.However,this consistency model would be damaged by image splicing or composing operation.Therefore,if the inconsistency of the pixel-related pattern is detected in the image,the trace of the splicing/ composing forgery can be found.According to this principle,we used the Spearman Correlation Coefficient to estimate the CFA interpolation correlation of test image,and compare the estimated image mode and the tested image mode,seek the inconsistencies.Then we use Canny operator to label splicing/composing image regions.The experimental results show that the proposed method can detect splicing regions accurately,and has good robustness and stability;its performance is superior to other algorithms that have the same function..An image splicing/composing detection method based on hyperpixel segmentation is proposed.Considering the spliced image is generally generated by using the images that are taken via different imaging equipment,thus,in the splicedimage,the different regions often have different correlation.Based on this idea,we use hyperpixel segmentation theory,combine with the classical K-SVD denoising dictionary learning algorithm,and establish a method of image spliced regions detection.In the proposed method,we first use K-SVD(Kmeans-Singular Value Decomposition)dictionary learning denoising algorithm for image denoising,and use the SLIC(Simple Linear Iterative Clustering)algorithm for super pixel segmentation.Then we extract the correlation features;define criteria according to the values of the correlation features,and then the splicing region detection algorithm is established.Experimental results show that the proposed algorithm can accurately detect splicing image regions,and the detection accuracy achieves 99.54%.It is obviously that the proposed method is superior to the existing splicing region detection algorithms.