The Theory Of Digital Watermarking Algorithm Based On Wavelet Analysis

In this paper, we introduced the concept of digital watermarking, includingdefinition, classification and etc. With the appearance of E-commercial and Internet, itis getting more and more important to protect the copyright of picture, soundtrack andother multimedia resource.Then, focused on image protection, we offered the concept of wavelet analysistheory. With the characteristic of MRA solution, wavelet is efficient and accuratedealing with the process of non-stationary signal, such as images. Especially in thefield of watermark algorithms, the theory of wavelet analysis offered irreplaceablestrength.Two algorithms were proposed in this paper. According to the Human VisualSystem, we offer a self-adapting way the set the depth of embedding, in which thedepth was calculated in the process, not fixed defined. Also, the watermark wasembedded twice, both in high-band and low-band. The other one we proposed is socalled non-embedding type. The watermark is not embedded into the source image, butrather, used to derive a secret key, which contains the image features and thewatermark. First of all, it guaranteed the invisibility, and then in the test part, we cansee this algorithm offered extremely strength against various attacks. Because of itsnon-embedding characteristic, some geometric attacks were also defeated. We usedbigger icon of 256 grey levels, with contains much more information than only bi-levelmarks.Beside, aimed at the disadvantages of traditional standard, such as PSNR, weproposed a test bench to evaluate the visual effect of attacked image and extractedwatermark in this paper. It turns out that the test bench is more effective at least in thispaper.At last, we pointed two topics to be developed in future: one is dealing withtrue-color images; the other is deciding best wavelet functions. We did preliminarytests on both subjects and got basic conclusions.