Research On Image Tampering Detection Technology Based On Perceptual Hash

With the rapid development of multimedia information processing technology and computer network,more and more multimedia resources can be obtained conveniently through the network.Digital images have become one of the most important ways to acquire and publish information in our daily life because of its intuitive and easy-tounderstand characteristics.However,people can easily edit images with the development of the image processing software.While digital images convenient for people,security issues such as the authenticity and integrity of digital images are exposed and become an urgent problem to be solved.Image perceptual hashing provides a more convenient way to solve the problems related to image authentication based on the understanding of image content.This paper firstly presents a literature review of image hashing for image authentication in the last decade,and classifies these algorithms into different categories.Secondly,a new perceptual hash learning algorithm scheme is proposed for solving the problems of existing schemes.Th specific research works are as follows:1.A Multi-View Semi-supervised Hashing algorithm with Perceptual Saliency(MV-SHPS)is proposed,which explores supervised information and multiple features into hashing learning simultaneously.Firstly,the algorithm converts the image from the RGB color space to the CIE LAB color space,which is more in line with the characteristics of human eye system,and extracts its L component to perform Int Wavelet Transform to extract its coefficient matrix for further processing operations.Secondly,the image multi-view features are extracted and merged by using the learned hash function which can generate a perceptual hashing with shorter length and better performance.Finally,in the stage of the hash similarity matching,the image saliency feature is computed and used to assist tamper detection to improve the overall performance of the algorithm.A large number of experiments have verified the good robustness and distinguishability of the proposed multi-view image feature embedding scheme.2.Based on the above the image multi-view feature embedding hashing learning algorithm,a binary multi-view perceptual hashing learning algorithm is proposed.Specifically,the proposed hashing learning algorithm mainly includes two core components: offline learning and online hashing.Offline learning is used to learn hash codes and simultaneously to generate a hash function.It mainly includes two stages,Collaborative Binary Representation Learning(CBRL)and Perceptual Content Authentication Learning(PCAL),which can merge image multi-view features into the common hamming space.For online hashing,the visual features of received image were extracted followed by mapping into binary codes with hash functions.Extensive experiments show that the proposed binary multi-view hashing algorithm can have good robust performance for multiple types of image content-preserving processing operations,including: additive noise,image filtering,blur,JPEG compression,illumination adjustment and so on.