The Research Of Digital Image Watermarking Based On Support Vector Machine And Genetic Algorithm

With the development of internet and the widespread use of computer, digitial image can be downloaded and modified easily from internet. Therefore, how to protect the owner's copyright of digital image has become an important issue. Digital image watermarking scheme is an appropriate way to solve this problem, and millions of researchers are now involving in the research of digital image watermarking.In digital image watermarking, watermark is embedded into original image by embedding algorithm. It will be extracted form digital image for copyright protection when the image is used illegally. However, the conflict between invisibility and robustness exists in digital image watermarking. The request of invisibility means the modification by embedding algorithm should be small, while the request of robustness requires large modification. This article is focus on finding ways to solve this conflict. Our works are introduced by follow:(1) A novel blind digital image watermarking scheme based on discrete cosine transform domain (DCT) is proposed. Experiments results show that the new scheme that is improved by Shieh's scheme has good fidelity and robustness.(2) This paper proposes a new digital image watermarking scheme based on human visual system (HVS) and support vector machine (SVM). After researching on several human visual systems that are used in digital image watermarking, the scheme constructs a classified model for digital image locations by considering brightness and texture. Then, support vector machine is used to simulate the classification model and classify image locations into several types that have different sensitivities to human eye. Different embedding strengths are selected to different types for gaining high fidelity and robustness.(3) This paper proposes an image watermarking algorithm based on support vector machine (SVM) and genetic algorithm (GA). In the scheme, support vector machine is used for optimizing embedding strength, and genetic is employed to optimize embedding location. By considering optimization of both fidelity and robust, the fidelity and robustness of our scheme are better than the scheme that be proposed in the past papers. Meanwhile, as the time that is consumed by genetic algorithm is so long that an improved genetic algorithm is proposed in our scheme for descending time consumption.Based on the researches, our schemes are simulated on the platform of Matlab 7.0. Two systems have been accomplished, one is based on support vector machine and the other is based on both support vector machine and genetic algorithm. The first is used for an immediate situation. Although the time that is comsumed by the other system is long, we will obtain better results than the first one. Experimental results show that our scheme has both high fidelity and good robustness.