| With the development of science and technology, digital image has been getting moreand more widely used and spread. However, the advent of image processing softwares whichare low-cost, high-performance lead that the digital image is vulnerable to a malicious forgery,thus there exist a great security risk. Once the tampered images are abused in scenarios suchas monitoring system, news report, conviction and so on, it may cause serious negative impacton economic development and social progress. So there is an urgent need for a technology toverify the authenticity of digital images.In this paper, forensic techniques for verifying image authenticity are investigated in twoaspects: tampering type identification and copy-paste forgery detection. The main work canbe summarized as follows:1. The typical algorithm of detecting tampered type based on image features and patternclassification was simulated. For the problem of how to detect the tampered type of imagesundergone copy-paste forgery, the corresponding improvement ideas were proposed.2. The scale invariant feature transform-based detection algorithm for copy-pasteforgeries proposed by Huang was simulated and analyzed, and then its deficiencies werereviewed. Based on this, a new type of feature and a new feature matching method wereproposed. The proposed algorithm reduces the time complexity while ensuring theperformance of the original algorithm.3. Because most of the existing copy-paste forgery detection algorithms are mainly at thecost of higher time complexity, we focus on the research of matching algorithm. First theexisting matching methods were discussed respectively, the advantages and disadvantageswere analyzed. Then a new method based on locality-sensitive hashing was proposed, thematching number and matching time in the case of81and225-dimensional feature vectorswere compared with Best-Bin-First. The new method can detect more similar vectors, andsupport high-dimensional data matching. Besides, time complexity is low. |
PDF Full Text Request