The Technology Study Of Picture Digital Watermarking Based On Wavelet Analysis And Neural Network

Digital watermarking technology is a new direction of the information security field,is a new technology to protect the digital copyright,to authenticate digital information sources and integrity in an open network environment.Watermarking algorithm is divided into spatial and transform domain algorithms according to the hidden location of the watermark.In contrast with transform domain watermarking algorithms,the advantage of spatial domain watermarking algorithms is computationally faster, and time complexity is relatively low,with a large amount of hiding information etc.Drawback of spatial domain watermarking algorithms is fragile digital watermarking,robustness is poor.In contrast with spatial domain watermarking algorithms,the advantage of transform domain watermarking algorithms is that embedded watermark signal energy in the transform domain can be distributed to all the pixels on the airspace,which will help improve digital watermark invisibility and robustness.The disadvantage of the general transform domain watermarking algorithms is larger in computation.Wavelet transform in digital watermarking technology is mainly reflected in the application of digital image watermarking.The basic idea is to decompose the multi-resolution image to generate different spatial and independent sub-band images,to use the local characteristics of wavelet coefficients to achieve watermark embedding in region of interest.Neural Networks in the application of digital watermarking technology can be divided into two aspects.On the one hand,the use of neural network in watermark embedding to classify images or generate adaptive watermark on images,which aims to improve the embedding strength and the image of the security truth degree.On the other hand, using neural network in watermark detection,which aims to improve the accuracy of watermark detection.In the context of wavelet analysis and artificial neural networks,this paper presents a watermarking algorithm based on wavelet packet transform and Hopfield networks.The main research of this paper as followings:(1) This paper briefly describes the basic theory of the wavelet analysis,artificial neural network and digital watermarking,and studies three kinds of digital watermarking algorithms.(2) This paper presents a watermarking algorithm based on wavelet packet transform and Hopfield networks.The watermarking algorithm uses wavelet packet transform to select the low-frequency to embed watermark, to achieve a specific frequency band to embed watermark,in order to resist the filtering attack;The watermarking algorithm uses discrete cosine transform to make the sub-image energy more concentrated, in order to be effective against spin attack;The watermarking algorithm uses Hopfield neural network to detect the watermark,not only improves the accuracy of watermark detection,but also enhances the watermark extraction.The results show that the algorithm can resist cutting attack and compression attack;can resist a certain amount of noise attacks and scaling attacks;for filtering attack and spin attack the algorithm has strong robustness.(3) This paper carries out six watermark attack experiments on the proposed watermarking algorithm and the other three watermarking algorithms.Experimental results show that the effect of the proposed watermarking algorithm is better than the other three watermarking algorithms against compression attack,noise attack,attack filtering,rotation attack.(4) This paper attempts to use particle swarm optimization embedding strength parameter in the proposed watermarking algorithm.In the experiment, found that only can optimize the parameters on the specific problems in specific areas and the optimized parameters are not universal.Optimization of parameters should do further research in the future.