Research On Digital Watermarking Techniques For Images

With the development of digital technology and internet, many kinds of digital multimedia are transported through Internet. Digital multimedia could be replicated and transported rapidly with low costs without permission. However these characteristics of digital multimedia transmitted over the Internet would be easily attacked by the pirates. It is very important to find new ways to protect the digital works against the attackers. Besides the traditional protection laws and administrative methods, it is important to protect the digital content utilizing their instinctive characteristics. So the digital watermarking technique is becoming one of the hot research projects around the world.In this paper we study the main features of the current techniques, and analyze the typical watermarking algorithms for the gray images and the color images.First, the watermarking techniques based on spatial domain and transformation domain for the gray images are studied. The LSB algorithm is implemented and the robustness of this algorithm is tested under several kinds of attacks. This algorithm could resist the zooming attack but it is not robust to clipping and other signal processing. Then an improved algorithm based on DCT transform is proposed. The watermark is embedded into the DC coefficients after the DCT transform not the AC components. The experimental results show that the technique could resist most of signal processing attacks. However, the watermarking payload of this method should be lower, otherwise the quality of the original image would be degraded. We also evaluate an adaptive blind image watermarking algorithm based on the block DCT transform. The watermark is a visible binary image. We use two independent pseudo-random sequences to substitute the 'O'andT in the binary image, and the original spectrum will be spread over a wide area. The modulated signal is embedded in the low-middle frequency bands of the block DCT transform, and the scheme can adaptively adjust the strength of the embedding signal according to the original magnitude and can extract the watermark without the original image. The simulation results demonstrate that the method is robust to most of the image procession operations such as lossy JPEG compression, low pass filtering, median filtering, cropping etc.Second, we analyze the watermarking algorithms based on spatial domain and transformation domain for color images. Firstly, we transform the watermark signal into binary data stream and embedinto the least significant bit of every byte in RGB color space. The capacity of the watermark image scales the size of the original image. This method holds high data hiding rate. It is robust to most of normal signal processing attacks. A valid blind watermarking algorithm is introduced here. The images that have a lot of white pixels usually can produce a lot of overflows after the watermark is embedded. The algorithm embeds the watermarking signal modulated on some fix positions which keep a certain distance from each other. Some decimals are added to the image-signed and different parameters are embedded for different images. The experiments show that this algorithm is transparent, robust, accurate and cost-efficient.Third, this paper studies two watermarking algorithms based on the wavelet domain. One is to embed the watermark sequence into LH and HL coefficients of wavelet transforms. We use two independent pseudo-random sequences to substitute the 'O'andT in the binary image, which has the same effect of the spread spectrum. So this algorithm has a better quality of robustness and a better quality of resisting the damage from clipping, noise, JPEG compression, modification and shifting. Another is to embed a 24-bit color image into the HH coefficient of another 24-bit color image. Color spaces with linear relation to RGB color space with uncorrelated components are found to be most suitable for watermarking application. The YUV color space is more suitable to embed watermark. We first transform the RGB color space to the YUV color space. Then we apply wavelet transformation to the respective YUV components of the original image and watermark image. The YUV components of the watermark image are embedded into the corresponding YUV components of the original image. The future research work is to find the more effective embedding location in the original image.