Study On Digital Image Watermarking Algorithm Based On Wavelet Transformation

With the rapid development of computer and network technology, the information of digital multimedia is becoming more easily to acquire, while at the same time, it also more easily suffers the illegal copying, fabricating, tampering, infringement etc. attack. The conventional safety precautions which based the cryptology have the ability not equal to their ambition. In this kind of situation, the digital watermark technology appears. Digital watermarking can embed the secret information into the carrier data, and check it out when necessary. It requires that the embedded data wouldn't affect the normal use of the carrier data, which makes the invisibility requirement of digital watermarking. Besides that, carrier data may be processed while transmitted before used, so digital watermarking should also be robust.Early research of digital watermarking dwelled on space field of images, but it is proved that it is not robust. It is proved that human vision has different sensitivity towards different parts of images. In order to improve the robustness of digital watermarking, the watermark should be embedded into the important visual parts of the images, such as low frequency part. It leads to the development of transformation field watermarking. Present researches mostly concentrate on transformation field watermarking.The theory of Wavelet Transformation is a new math branch which emerged in recent years. It is a landmark since Fourier put forward Fourier Transformation in 1822, and can solve many problems which are hard for Fourier Transformation. Although Fourier Transformation has been widely used in signal processing field and can describe the frequency attributes of signals well, it meets problems while processing abrupt and non-stationary signals. Wavelet transformation can be treated as the extension of Fourier transformation, which means it is a local transformation between space (time) and frequency. The general idea of the use of Wavelet transformation in image processing is to perform multi-resolution decomposition with the target image, and then processing the coefficients of the sub-images in different space and frequency. The core is the processing of the coefficients, such as image compress and coding. According to the tower decomposition put forward by S .Mallat, the target image can be divided into 4 frequency bands: horizontal, vertical, diagonal and the low frequency part, and the low frequency part can still be further divided. Images generated by wavelet decomposition have different attributes to the original image. The power of the original image concentrates on the low frequency part, which is the most important part to human vision, which there's much less power in the detail parts of the image. The high frequency parts present the horizontal, vertical and diagonal edge information of the original image. Because of the development and the characteristic of the wavelet transformation, this article emphatically carries on the wavelet transformation field watermarking research.First of all, the thesis carries on the systemic comparatively research of the algorithm of digital watermarking which is based on wavelet transformation. The background and the recent development of the digital watermarking is introduced, then dissertate the digital watermarking about the basic concept, generic model, typical algorithms, attack analysis and evaluation of performance. The thesis also introduces the theory and application of wavelet transformation, the principle of image permutation.Then the algorithm is proposed which is used to embed watermarking into original image several times in wavelet field to improve the robustness of the watermarking. The algorithm is done by taking the same number of layers wavelet decomposition for both watermarking and the original image. Then the original image's high frequency and low frequency coefficients are broken into pieces for each resolution, in which the all pieces are the same size of the watermarking. After that, embed the watermarking into the every piece of the original image. As talks to the drawing, adding the each piece of high frequency and low frequency coefficients for the same resolution and then give the mean value of them. The simulation results say that the robustness of that algorithm is good in some way, however, it is not very good at resisting the cutting attack. In this thesis a new draw algorithm is proposed for resisting the cutting attack. For this algorithm there is no need giving the mean value when drawing. The simulation results say that the algorithm can draw the watermarking effectively even though the image has been cutting greatly. There is no one algorithm can resist the various attacks wholly. It is because that the drawn means are oneness, it might be solved by taking extensive algorithm to resist the various attacks.At last, a digital watermarking algorithm which is based on the discrete wavelet transformation is proposed under the analysis of the transformation field digital algorithm. The thesis is arranged as follows:(1)Making the binary image as the watermarking information, and then pre-treat the watermarking information before it is embedded. The chaos sequence is produced according to the chebyshev mapping. Because the over sampled chebyshev has one more modulation parameter, the sequence produced by over sampled chebyshev is safer. Then do chaos treatment for the watermarking, making them into disordered information. The safety and uncertain of the watermarking system could be enhanced for the characteristic of the chaos sequence.(2)Take the two-layer wavelet analysis for the original image according to the characteristic of the wavelet transformation. Then embed the watermarking information into the HH2 wavelet coefficient.According to the HVS, eyes are not sensitive to the change of the diagonal part, so after embedding the watermarking information into the HH2 coefficient, the watermarking is also invisible. When the embedding is done, divide the HH2 coefficient into some parts and the number of parts should be the same as the size of the watermarking. Subtract the mean value of a particular part from each HH2 of the part so that it can get a lot of embedding information and assure the robustness of the watermarking.(3)Create two sequences according to the Logistic chaos Mapping,and the two sequences must have little relativity. When the pixel of the watermarking is 0 or 1, embed different sequence values. Take the corresponding calculation for the HH2 value of wavelet analysis and the two sequences when drawing the watermarking. (4)The simulation results say that the algorithm has a good robustness for image processing at the same time it can assure the watermarking is invisible. Then make the image that has been embedded the watermarking to suffer the noise, compress, filter, and geometric attacks. By lots of simulations, it can prove that this algorithm has a severely robustness for geometric attack.. Especially when the image that has been embedded the watermarking is 4 times smaller that before, the watermarking can not be drawn because of losing too much information.The algorithm is easy for embedding and drawing; the information of the watermarking embedding is large, moreover, the watermarking can been drawn blindly. As a whole the algorithm is the first generation algorithm, so it is necessary to ameliorate the algorithm by the idea of the second generation watermarking technology.