Fragile Watermarking Algorithm Based On Image Characteristics Of High-security Research

In some image applications filed, it is important to verify that the image has not been manipulated and that the image was originated by either a specific camera of a specific user. Fragile watermarking which provides effective solutions for image content authentication and tamper-detection is exactly such a technique.First of all, we investigated the digital watermarking and authentication digital watermarking theories systematically, summarized the general techniques and general models of them. Fragile digital watermarking techniques were analyzed in detail, pointed out its characteristics and designing requirement and limitations. At the same time we analyzed typical algorithm of various kinds and main attack method of the fragile watermarking.In the next place, two new fragile watermarking algorithms with high security and image's feature are proposed. The details are illustrated as follows:1. A spatial-domain fragile watermarking based on global feature and discrete cosine transformation is proposed. In the first stage, this algorithm takes the approximate sub-band computed by two level discrete wavelet decomposition as the global feature, with the key, which is the initial value of Logistic mapping. And then, it generates a random matrix by which the LSB of the original image is set to zero. After being set, the image is transformed into 8×8 blocks using DCT. Being quantified, the high frequency coefficients of DCT are taken as the information generating the watermarking. The algorithm can find subtle tamper in spatial domain by use of the sensitive of the high frequency coefficients of DCT. By scrambling the coefficients, it is also able to increase the security of the algorithm.2. Combining the advantages in algorithms of spatial domain and frequency domain, a new algorithm in combinational domain based on is proposed, which is of high location accuracy. In the algorithm, the original image is randomly divided into blocks in both spatial domain and frequency domain. The watermark information in spatial domain is embedded into the DCT domain. So the same information is embedded three times. After that, the watermark block in frequency domain's watermark is embedded into the LSB in spatial domain. The algorithm has the less watermark information capacity and advantages of watermarking both in spatial domain and in frequency domain, and the precision of locating the tampered image block.In this thesis, these algorithms are implemented using MATLAB and tested with a large number of practical data. The results demonstrate that both of watermarking algorithms are of characteristics of fragile watermarking, strong location ability, high security and practicability.