| With the development of digital technique and Internet, digital media can be copied and transported with low cost and high speed, which makes it convenient for both media creators and users, and challenges the protection of media copy write as well. Researchers have been searching ways to protect digital media and digital watermark is one of the hottest techniques. This paper mainly concerns digital image watermark and tries to bring up new color image watermark algorithm and blind watermark extraction model as following:First, the paper introduces the digital watermark's concept, characteristic, classification typical algorithms, main application areas and varies of attacks.Then, two typical watermark algorithms, the LSB based and the DCT based, are analyzed and implemented with MATLAB.Next, a new color image watermark algorithm based on image block average value is brought forward. By dividing an image into many sub-image blocks, constructing the images steady features with block average value modulation, and modulating and recognizing the fitting region of each block's blue channel average value, the watermark is embedded and extracted. This algorithm is robust to image region processing attacks.At last, an improved blind watermark extraction model is designed, which tries to solve the problem of low extraction efficiency of traditional watermark models. The new model thinks of the embedded watermark as a kind of noise and before watermark extraction, it makes use of noise-decreasing methods to get the estimation of original image. To compare the estimation to the image to be detected, the noise containing watermark is gotten. The noise is then amplified and watermark can be gotten back with partial maximum code principle on the amplified noise. Experimental results show that the model, when used in frequency domain blind watermark algorithm, can solve the problem of watermark extraction arising from interpolation and multi-transformation errors. |
PDF Full Text Request