Abstract : Digital image watermarking has attracted a lot of interest in recent years, due in particular to the development of Internet and the World Wide Web. The aim is to protect ownership by including in the image a copyright information. This information, or mark, has to be set in such a way that it is invisible: indeed, it must not alter the viewing content and, in addition, it should not be easy to remove. Furthermore, the mark must be resistant to attacks directed at erasing it. A number of methods have been proposed to insert robust and invisible watermarks. Some operate directly in pixel space, other in a transform domain, such as Fourier or DCT . We propose here to study a wavelet packet based watermarking procedure. Working in the wavelet domain yields a number of advantages: First, it allows to control in a precise way the location both in space and scale of the mark. Second,
wavelet coefficients give a structured way of representing the information: as is well known for instance from studies in image compression, in most cases, only a few coefficients are large, indicating where the information lies in a given image in terms of scale and space. This is useful both for invisibility and robustness concerns. Using wavelet packets adds another degree of freedom because it allows to select frequency independently of scale. Finally, wavelet based algorithms are fast and allow to reconstruct the image. After an overview of the algorithms recently used, we recall some basics facts about wavelet packets decomposition and best bases selection criteria. Thirdly, we presented our watermarking algorithm. The part 4 gives amelioration of results under attacks and the last part presents an analysis of the proposed method.