1 January 2007 Near-lossless spread spectrum watermarking for multispectral remote sensing images
Author Affiliations +
J. of Applied Remote Sensing, 1(1), 013501 (2007). doi:10.1117/1.2535355
Watermarking represents a potentially effective tool for the protection and verification of ownership rights in remotely sensed imagery. Such data however cannot undergo critical quality degradation processes, for they must ensure accurate extraction of thematic and scientific information. In this paper, we propose to adapt a state-of-the-art spread spectrum method found to be particularly powerful for real world images to make the watermark insertion process near-lossless while preserving most of its effective robustness capability. This is done by embedding the watermark in the middle-frequency range of the discrete cosine transform (DCT) domain instead of the low-frequency range. The exact position is determined by a numerical root-finding method, targeted to achieve a userspecified minimum level of robustness against a given ensemble of attacks. In this way, an optimal trade-off can be obtained between quality and robustness. Thorough experimental tests over a multispectral remote sensing image show that the proposed method allows to trade a contained loss in robustness with a significant gain in image quality, quantified in terms of both peak signal-to-noise ratio (PSNR) and impact on classification performance. In addition, they reveal an interesting property of the watermark insertion shifting into the middle-frequency range, consisting of a sharp increase in robustness to the widely used cropping operation, which is considered the most critical attack in remote sensing imagery.
Farid Melgani, Redha Benzid, Francesco G.B. De Natale, "Near-lossless spread spectrum watermarking for multispectral remote sensing images," Journal of Applied Remote Sensing 1(1), 013501 (1 January 2007). http://dx.doi.org/10.1117/1.2535355

Digital watermarking

Image quality

Remote sensing

Multispectral imaging

Image classification

Discrete wavelet transforms

Image processing


Back to Top