Study Of Image Watermark Scheme Based On Smooth Support Vector Machine

Digital watermarking technology has become a hotspot in the field of multimediainformation security. It embeds some digital watermark that mark the specific meaning intothe digital audio, images, videos, documents and other media files, to prove the copyrightownership or tracking the violations. Meanwhile it can ensure the integrality and the reliabilityof digital information through detecting and analysis the watermark, thus becoming aneffective method of protection of intellectual property rights of digital media products.Support Vector Machine is a new technology in data mining. It was initially proposed byVapnik、Chervonenkis of the Soviet Union in1990s. SVM build on the strict theoretical basis,which make it better solve "the curse of dimensionality, local minima," over learning"problems. In recent years, SVM theory has been widely applied in many areas. Pa rticularly,some experts at home or abroad try to introduce SVM to digital watermarking in order toimprove the performance of watermarking system through utilizing the excellent features ofSVM. From the experimental results, most get certain effects. However, there are still thefollowing main problems: First, the SVM-based watermarking method is basically non-blindwatermarking. Is there a better way to combine the SVM with watermark technology to getblind-watermark remains to be further research; Second, the form of SVM-based watermark istoo single. Most relative paper utilize support vector classification machine to determine theembedding location and intensity of the watermark or support vector regression machine toobtain the embedded watermark information; Thirdly, SVM reduce the efficiency of theimplementation of the watermark algorithm in a certain degree.This paper inspired by the application of the standard SVM-based watermarkingtechnology further research on the SSVM-based watermarking technology. First researchingthe model of the standard SVM and the digital watermarking technology, and then combinedwith the result of the standard SVM-based watermarking technology. Finally, establish theSSVM-based digital image watermarking model and propose an SSVM-based imagewatermarking scheme in the wavelet domain. The experiments say that this method not onlyachieves satisfactory results of the invisibility, robustness, but its efficiency also much higherthan the SVM-based watermarking technology. Therefore, this paper on one hand proposes anew solution for the problem of digital image watermark, on the other hand makes a usefulexploration for the support vector machine theory and application development.