Design and analysis of digital image watermarking using wavelet transform

Abstract

Digital watermarking is a process of embedding data called watermark into a multimedia object such that it can be detected or extracted later to make an assertion about the object. It has been used in numerous applications and on a variety of host signals. The applications include authentication, fingerprinting, broadcast monitoring, copy protection and meta-data tagging. The host signals include images, audio signals, video, printed documents, maps, computer programs, animation parameters and 3-D models. Of the various watermarking algorithms, some outperform others in terms of basic watermarking requirements like robustness, invisibility, processing, costs and security. In this thesis, we designed and analyzed a new image watermarking system in the wavelet domain for single and dual watermark embedding. A meaningful and recognizable image has been used as the watermark which immediately provides strong authentication information instead of using the random number generator to create the watermark. In addition, to achieve the security aspect of the watermark, the bits of the watermark image are randomly permuted so that only those who have the key can reconstruct the watermark. To further enhance the security of the watermarking algorithm, we propose the watermark be divided into small pieces and these different pieces are embedded in different locations of the cover image. Each piece needs a separate key to generate PN sequence which is used to encode the watermark bits. This provides secrecy of the watermark against public extraction and detection since without all the right keys, it will be impossible to decode the watermark. For single watermark embedding, three different techniques of wavelet transform, namely, Discrete Wavelet, Wavelet Packet and shuffled wavelet coefficients are proposed. Dual watermark embedding is developed based on the single wavelet embedding to embed two watermarks simultaneously in the same image space. Six combinations based on the three different techniques of wavelet transform proposed in single embedding have been experimented. The proposed technique of hiding dual watermarks can simultaneously achieve source authentication and fingerprinting, where one watermark determines the sender and the other watermark determines the original recipient. It may also be used to resolve deadlock copyright protection problems as outline by Swanson et al.(1999). As opposed to the existing techniques where the watermark is directly embedded into the wavelet coefficients, our proposed single embedding method requires the wavelet coefficients of the host image to be shuffled before embedding the watermark. Hence, a higher security is achieved from our proposed technique since a secret key is needed to shuffle the wavelet coefficients. As for our dual proposed watermark embedding, the watermarks are embedded at different sub-bands of wavelet transform using different techniques as compared to published works where both the watermarks use the same embedding technique(Lie et al.,2003; Tao & Eskicioglu, 2004) or using different domain(Chaw-Seng et al.,2005). Our experimental results show that embedding location, gain factor, wavelet function and detection threshold strongly influence the quality of the watermarked image and the extracted watermark. The proposed single watermark and dual watermark based embedding methods fulfilled the requirements of a watermarking system in terms of imperceptibility, robustness, verifiability of the extracted watermark and security.