Research Of Digital Watermarking Algorithms Based On Block Replacement

Digital watermarking is a kind of digital signal or pattern which is identifiable and can be embedded into the host signals. Its embedment does not affect the availability of the host signal. Digital watermarking technology can protect the copyrights of some kind of multimedia digital works (images, video, audio, etc.). This paper introduces the concept, principle, system framework and evaluation criteria of the digital watermarking technology. Completion process of watermarking (embedding and extraction test) and the robustness test (anti-Attack Analysis) were researched. The block-classification methods and replacement technology of the digital watermark were also analyzed. On this basis, Combining with the characteristics of theHuman Vision System(HVS) characteristics, a self-adaptive digital watermark algorithm based on block replacement is proposed.In the algorithm, the host image is divided into blocks firstly. Then, according with the masking effect of the human visual system, the image blocks are classified into smooth blocks, texture blocks and edge blocks by using the method of variance and edge detection, thus achieving the adaptive strength of the watermark embedded based on classified weights. Secondly, the algorithm divides the watermark image, completes embedding based on self-adaptive intensity when watermark block and host image block obtain the best similarity. At the same time, a position matrix is generated to record the embedding location of the host image. The watermark extraction is the inverse process of watermark embedding. After obtaining the locations of the embedded watermark from the position matrix, the watermark blocks are extracted according to the watermark embedding strength of its classification. The similarity functions used in this algorithm adopts dot product, Nprod (Normalized product-relation) and MAD (Mean Absolute Difference). Some texture and glossy images are used in simulation experiment, the results show that the algorithm has good invisibility for gray images and color images, and it is robust to some common attacks.