Research On Passive Forensics Technology Of Tampering With The Same Image

The prevalence of various image editing software has made image tampering more and more convenient.Image forensics technology is used to detect whether the image has been tampered with.Active evidence collection requires the implantation of digital signatures or digital watermarks in the image.Passive forensics only needs to be detected,so passive forensics has become a hot research.Tampering of the same image is more common,because the areas in the same image have the same illumination and shadow,making it difficult for the naked eye to identify the tampering area.A lot of literatures have proposed their own detection schemes in this field,which are mainly divided into detection techniques based on image block extraction features and detection techniques based on key point extraction features.Image block-based detection techniques often divide images into overlapping blocks of fixed size,so there are problems such as high detection complexity.While key-point based techniques,the speed at which key feature features are extracted is greatly enhanced.However,the key points are often sparsely distributed and face the shortcomings of incomplete detection.Aiming at the shortcomings of key points-based detection and incomplete tampering,a segmentation-based copy-and-paste tamper detection algorithm is proposed.The algorithm extracts the SIFT key points first,then obtains the matching image blocks according to the g2NN matching algorithm,and then uses the custom clustering algorithm to remove the mismatch.The clustering algorithm significantly improves the removal of the mismatched regions,and finally uses adaptive segmentation.The size of the image block is changed according to the picture.The SLIC segmentation divides the image into non-overlapping blocks,and marks the tamper region according to the key points matched earlier.The algorithm more accurately indicates the tampering region.The experimental results show that the proposed algorithm not only can detect single-copy and multi-copy tampering maps,but also recognize tampering maps with rotation and scaling geometric transformations.It is also robust to JPEG compression,and the detection results are more accurate,with few mismatches region.Aiming at the problem that the traditional image block-based algorithm has large memory consumption and is not robust to post-processing,a detection algorithm combining SURF and Zernike moments is proposed.The SURF feature detection has the advantages of simple operation,high efficiency,and short calculation time,and can quickly locate the copy and paste area of an image.Then,an image block is obtained centering on the key points obtained by SURF matching,and the image blocks are overlapped and divided,and then the Zernike moment of the image block is extracted to represent the image block features in detail,and finally the local sensitive hash is used for matching.The experimental results show that the proposed algorithm effectively reduces the number of image blocks and is robust to tampering images with rotation operations.It can detect tampering images after JPEG compression and has good effects on Gaussian blurred images.