Research On Methods Of Perceptual Hashing Content Authentication For Color Image

As the most intuitive and direct approach for people to obtain information, image has teemed in various fields of society in the new era of digital multi-media. Considering that image is easy to edit and modify, with the increasing number of image processing software, more and more people question the security and integrity of image. Image perceptual hashing can map image which has same content into a brief summary, so it can be used to image content authenticity and integrity authentication. This thesis starts with color image in practical application, emphasizes on not only authenticity and integrity of color image, but also tampers detection and location by perceptual hashing technology. The detail researches are as follows:1. A color image perceptual hashing algorithm based on non-subsampled contourlet transform is proposed. Firstly, the algorithm adopts non-subsampled contourlet transform combined with image normalization. And the statistics obtained from the transform coefficients are regarded as the intermediate hashes. Finally, it makes use of the relationship of statistic comes from different blocks to quantify the intermediate hashes. This algorithm encrypts the quantified hashes with chaotic sequences to compose the final hashes. The experiment results illustrate that the proposed algorithm is strongly robust to common content-preserving modifications and has good uniqueness. Furthermore, the proposed algorithm is better than compared algorithms in the collision rate and error detection rate, so it can be widely used in related applications of image identification or authentication.2. Considering that existing algorithms can not include all characteristic content of the image, that is to say, they ignore the content of four corners outside the inscribed circle, we propose a color image perceptual hashing algorithm using interpolation mapping and non-negative matrix factorization. Firstly, the algorithm makes use of interpolation mapping the content of the four corners of a square image into a circular one, so it retains the image information as much as possible. Secondly, the algorithm divides the circular image into rings to make it resistance to rotation. Finally, this algorithm makes use of the NMF to transfer the quadratic mapping image from the rings to form the final hash values. The experimental results illustrate that the proposed algorithm not only has a good discrimination performance and be robust to most image content preserving operations, but also has a good ability of tampering detect, it can detect the content tampering of the image edges effectively and resist arbitrary rotation attack.3. In order to detect and locate tampered region in image more accurately, an image perceptual hashing authentication algorithm which combines NSCT feature and FT(frequency-turned, FT) feature is proposed. The algorithm simulates similar human perception, extracts salient regions by bottom-up way as local feature, uses statistical properties of NSCT as global feature and combines the global features with local features for image authentication. The global feature are used for distinguishing whether image is different image or not, and local feature are used to detect whether this image is tampered or not and locate tamper regions. The experiment results illustrate that the proposed algorithm has good robustness and discriminative. And it can detect tamper effectively and achieve more accurate tamper localization.