Academic journal article Journal of Management Information and Decision Sciences

Image Watermarking by Wavelet Decomposition

Academic journal article Journal of Management Information and Decision Sciences

Image Watermarking by Wavelet Decomposition

Article excerpt


The effective protection of intellectual property rights is of priority concern in the light of the accelerated growth of digital media. Digital watermarking has been increasingly recognized as a viable means of protecting the intellectual property rights associated with multimedia data. To fulfill this requirement, various watermarking schemes and software products have been developed (Hartung & Kutter, 1999).

Various investigations have developed digital watermarking schemes for images by Walton (1995), Bender et al. (1995), Bruyndonckx et al. (1995), Koch & Zhao (1995), Cox et al. (1997), Swanson et. al. (1996), Hsu & Wu (1998, 1999) and Memon & Wong (1998). In order to identify the multimedia data-related copyrights, the watermark should have the following features: imperceptibility, undeletable, statistically undetectable, robust to lossy data compression, robust to signal manipulation and processing operations, and unambiguous. In most related works, such as the works of Walton (1995), Bender et al. (1995), Bruyndonckx et al. (1995), Koch & Zhao (1995), Cox et al. (1997) and Swanson et. al. 1(996), the watermark is a symbol or an ID number which comprises of a sequence of bits. Only quantitative measurements can verify the detected results. Therefore, an experimental threshold is chosen and compared with the quantitative measurement to determine whether the host image is watermarked. Ambiguity can happen when the quantitative measurement estimated from an attacked image is in the proximity of the given threshold.

In this work, we present a novel image watermarking method based on wavelet decomposition, where the watermark is a visually recognizable pattern. Hence, a visual pattern can be unambiguously extracted for subjective measurement. In addition, a similarity measurement is taken for objective test as well.

Wavelet theory has emerged as an effect means of representing image signals in terms of a multiresolution structure (Daubechies, 1992). Based on multiresolution representation (Akansu & Haddad, 1992; Vetterli & Kovacevic, 1995), a signal is divided into a number of components, each corresponding to different frequency bands. Since each component has a better frequency and time localization, the multiresolution decomposed signal can be processed much more easily than its original representation. Multiresolution techniques are characterized by their successive approximation property. Restated, as higher frequencies are added, the higher resolution images are obtained. The theory of binary wavelet decomposition has been developed by Swanson & Tewfik (1996). This binary wavelet transform uses simple modulo-2 operations, and also shares many important characteristics of the real-field wavelet transform. The intermediate and transformed results are all in binary form, therefore no precision error is introduced and the decomposition is completely invertible.

A multiresolution watermarking technique can be constructed on the basis of the multiresolution structures of wavelet decomposition, both on real field and a binary one. In the proposed method, a gray-level image is decomposed into a multiresolution representation by real-field wavelet transform. On the other hand, a binary watermark is decomposed into another binary multiresolution representation by binary wavelet transform, which is invertible (or lossless). Since the Human Visual System (HVS) inherently performs multiresolution decomposition, each decomposed layer of a binary watermark is embedded into the corresponding layer of an image (Notably, to achieve imperceptibility, the lowest band of an image is left unmodified). Therefore, in cases involving attacks or progressive transmission, the coarser approximation of a watermark is preserved in the coarser version of an image. In progressive transmission, as higher frequencies are added, the higher-resolution image is obtained. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.