Image Perceptual Hashing Research And Application

With the rapid development of digital multimedia technology and Internet, multimediainformation has become very easy to copy, edit and spread. While people enjoy the convenience,the security and authenticity of multimedia information must be taken into account. Image as theprimary means of transmission of multimedia information, its authenticity and integrity of thecertification has become extremely important. Image perceptual hashing technology, as apromising emerging technology, provides a safe and effective solution for identification andauthentication of image content. The research of image perceptual hashing technology is farfrom mature, and many of the basic definitions and key issues are not fully resolved.We introduce the background and significance of image perceptual hashing technology andthe status of domestic and international studies. Based on the analysis of existing algorithms, wefound that the current common shortcoming of most algorithms is unable to resist the large-anglerotation attacks. In this paper, we studied the following topics:(1) Image normalization technique and based on this an image perceptual hashing algorithmis proposed, which can resist a large-angle rotation attack. This paper introduces the theory ofmoments and it is used to improve the existing image normalization techniques combined withthe theory of chaotic encryption. To evaluate the performance of the algorithm, the proposedalgorithm is evaluated and compared with other existed algorithms in following features: thehistogram, FAR curve, FRR curves, ROC curves, DET curves. Experimental results show thatthe proposed algorithm can be very good resistance to large-angle rotation, scaling andtranslation attacks, with good uniqueness, security and robustness against other comparedalgorithms.(2) We used Radon transform and the improved image normalization technique to constructan image perceptual hashing algorithm. We uses the properties of Radon transform withtranslation, rotation and scaling invariance, extract features from Radon projection curve andgenerates image perceptual hashing. Experiments show that the algorithm is able to resistlarge-angle rotation attacks, with good uniqueness, security and robustness against othercompared algorithms. (3) An image perceptual hasing scheme are designed, which used to the actual situation. Weused the two proposed algorithm to deal with the scheme and got the experimental results,compared the strengths and weaknesses of the two algorithms in practical applicationsAt last, we summarized the article and the outlook for the develop trend in image perceptualhash is given based on our study.