Research On Perceptual Hashing Based On Image Content Security

Hash,as a short sequence of response image content,has broad prospects for development and practical research.Robustness and uniqueness are key indicators for evaluating hashing methods.Image data is also faced with potential safety problems while it is growing in mass.Image data,as the most widely disseminated and most expressive type of information in multimedia data,is also the most vulnerable to attack.Some illegal workers are willing to pay the cost of breaking the law or Directly tamper with the image data or intercept the image directly during image data transmission and modify it with an image editing software to attack or tamper with it.As a result,the authenticity and integrity of the image are damaged,and even the content is distorted and can not be identified.The traditional image hashing method can only be effective against a certain type of geometric attack,and does not have robustness to other geometric attacks.This may result in image matching or authentication failure and the image security needs to be further strengthened.This paper starts with the feature extraction to study two hash methods with practical application prospects.Traditional image hashing methods can only resist one kind of attack,but resist weakly to many kinds of attacks,which has great influence on image authentication and image detection.In this paper,we present a method of hashing on local features of images.The method can describe the main content of the image,and can effectively extract the features of edge,corner and bright spot of the image,taking into account the unification of local and global contents.The scale-invariant feature transform(SIFT)algorithm can extract the 128-dimensional feature description vector of the image.Then the local preserving projection(LPP)algorithm is used to reduce the dimensions of the description vector to generate a hash sequence.Experiments show that the Hash method is robust and can effectively resist all kinds of geometric attacks with high image recognition rate.Aiming at the problem of too many feature points,complicated computation and low matching accuracy of traditional feature algorithms,an image hashing method based on Fast Robust Feature(SURF)and Principal Component Analysis(PCA)is proposed in this paper.The SURF algorithm has faster computation time,Better performance.In order to reduce the computational complexity,the SURF algorithm is adjusted to reduce the number of low-quality feature points by increasing the number of comparison points in the scale space.The SURF algorithm is used to extract the 64-dimensional feature descriptor of the image.Then use PCA algorithm to reduce its dimension to generate hash sequence.Experiments show that the Hash method has high recognition accuracy,robustness and uniqueness,which can meet the security requirements of images.