With the rapid expansion of World Wide Web, there is an increased need for multimedia information such as digital image, video, and speech. Because of this rapid increase in information, copyright protection of multimedia contents becomes an important topic. Watermarking is one technique that embeds watermark signals in multimedia contents so as to identify the ownership....[ Read more ]

With the rapid expansion of World Wide Web, there is an increased need for multimedia information such as digital image, video, and speech. Because of this rapid increase in information, copyright protection of multimedia contents becomes an important topic. Watermarking is one technique that embeds watermark signals in multimedia contents so as to identify the ownership.

In traditional watermarking schemes, watermark signals usually are pseudo random numbers, Gaussian noise, or 1-bit binary images. In this thesis, four watermarking schemes that use gray-scale logo images as watermarks are proposed. Two general features of these schemes are: (1) they require the original image to recover the watermark logo image and (2) watermarking is done in the wavelet domain.

In the first watermarking scheme, halftoning and inverse halftoning are used to convert gray-scale logo images into a binary sequence and vice versa. In the second scheme, a different bit allocation of wavelet coefficients for different subbands are chosen so that the embedded signal will not exceed the capacity of the original image, while the wavelet coefficients of the watermark signal are converted into 1-bit sequence before insertion. In the third scheme, a block-based switching method is proposed, where the original image and watermark are both divided into blocks in the wavelet domain and we then switch the blocks of the watermark so that the energy spectrum of the original image and the watermark are the same. In the last scheme, the embedding process depends on a statistical model of the wavelet coefficients of the original image. Experiment results show that all proposed methods yield good visual quality and are robust to a number of attacks such as compression and filtering. Our results show that in the future further research can be conducted using gray-scale logo to be the watermark.