Research On Blind Detection Image Digital Watermarking Technology

With the development of computer and network technology, and especially the rapid growth of Internet users and the startup of electronic business, it is a tendency for publishing and conveying digital products on network. However, The digital products are liable to be infringed and tampered in the procedure of publishing and conveying. Although network provides many conveniences for human, it renders us new researching subject: content reliability protection and copyright protection. Digital watermark is an important technology to the subject.Watermarking algorithm, which includes watermark embedding and watermark extraction or detection, can be distinguished as blind detection watermarking and non-blind detection watermarking. It is a blind detection watermarking if the original product is not used in watermark extraction or detection. This paper, taking still images as example, make deep researches toward blind detection semi-fragile watermarks and blind detection robust watermarks according to two principal application, that is authentication and copyright protection. Such technologies as quantization, reference watermarks and template watermarks are proposed and applied in the paper. The watermarks embedding does not adopt common embedding model, which avoid the model attack. The main ideas are described as follows:(1) Proposes a new semi-fragile watermarking algorithm for image authentication. The algorithm adopts biorthonomal wavelet for watermarks embedding, designs a quantization function, by which watermarks are embedding in biorthonomal wavelet transform coefficients. The paper studies chaos sequence and chaos key, and locates watermarking positions by chaos keys. Watermarks are embedded in all resolution-level, which enables multi-resolution detection. Various sensitivities of preventing tamper can be achieved by modifying quantization coefficient and detection threshold. Simulation results indicate the proposed algorithm can tell not only if the image is tampered, where it is tampered and tamper extent, but also the tamper type in light of multiresolution-level tamper assessment values.(2) Proposes a new lossy robust spread spectrum watermark algorithm. It ishard for traditional spread spectrum watermark to be blind detection. This paper borrows the idea from spread spectrum watermark, quantizes the wavelet transform coefficients the copyright protected image to produce a binary sequence, which is just the watermark. The binary sequence can also be produced by a chaos key. The method to produce the binary sequence is proposed in the paper. It is a blind detection watermark since we detect watermark by position key and chaos sequence key without original image. The watermarked image is lossy since inverse wavelet transform is not implemented, and then the wavelet coefficients are not modified.(3) Introduces Error-correcting coding from communication field to watermarking algorithm to improve robustness, which make the copyright information, such as user ID and stamp be extracted reliably. Bit plane decomposition technology is proposed for grey or colored image watermarks. The technology decomposes the grey or colored image watermarks into many bit planes, and the high-bit planes are coded with Error-correcting coding.(4) Proposes reference watermark to make watermark much more robust, especially to such local attacks as local blurring, cropping and replacement. A multi-channel watermarking model is constructed, which divides the wavelet domain into many channels. The robust and reference watermarks are embedded into all the channels. The reference watermark is employed to compute distortion ratios of various channels, by which we can estimate the robust watermark.(5) Finally, makes tentative research on robust watermark against geometrical attacks, proposes template watermarking algorithm. The algorithm embeds both robust and template watermarks into DFT domain. We estimate geometrical transform by template watermark and...