The Research Of Image Authentication Algorithm Based On Differential Energy Watermarking And Image Matching Algorithm Based On Hausdorff Distance

The rapid development of multimedia digital technology and Internet technology promotes the advance of the information society. Image matching and certification is emerging technologies in the field of computer vision and digital image processing. They plays an important role in the use of military navigation, medical image analysis and machine vision. At the same time, information security, integrity, certification of intellectual property rights and other issues also bother people. According to the type of images attack, this article makes a preliminary exploration and research in the field of image authentication algorithm and the image matching algorithm.(1) This paper studies the basic theoretical knowledge in image matching and image authentication. First, this paper summarizes the development history and basic concept of the image authentication and matching. Secondly, the paper introduces the image authentication and matching technology, including digital watermarking and image matching system model, the basic features and common image authentication and matching algorithm.(2)As for the noise, filtering and compression, corresponding image authentication algorithm based on Differential Energy Watermarking is proposed. This algorithm is based on the original increases the watermark scrambling and conditions. By selectively discarding some high-frequency DCT coefficients, the encrypted watermark is embedded in image. The experimental results show that the proposed algorithm can increase the amount of watermark embedding and resist the non-geometry transform attack. This algorithm is is a kind of high reliability, high safety of practical watermarking authentication algorithm. This algorithm is a kind of high reliability and safety of the practical watermarking algorithm.(3)As for the rotation, scaling, cropping in image matching, corresponding algorithm based on Hausdorff distance is proposed. The proposed algorithm extracts feature points by SIFT operator in original image and target image, and determines whether there is a kind of affine transformation between two images by an improved Hausdorff distance and box distance transformation. If there is an affine transformation which satisfies the requirement of the Hausdorff distance, the original image and target image can be matched. If there isn’t an affine transformation which satisfies the requirement of the Hausdorff distance, the original image and target image couldn’t be matched. The experimental results show that the algorithm can reduce affine space, improve the efficiency of image matching and eventually find the affine transformation between two images.