Secure Copyright Protection Techniques for Digital Images

Abstract

Abstract. This paper2 presents a new approach for the secure and robust copyright protection of digital images. A system for generating digital watermarks and for trading watermarked images is described. The system is based on a new watermarking technique, which is robust against image transformation techniques such as compression, rotation, translation, scaling and cropping. It uses modulation of the magnitude components in Fourier space to embed a watermark and an accompanying template and, during watermark extraction, reads a template in the log polar transform of the frequency domain. The template is used for analyzing scaling and rotation suffered by the watermarked stego-image. The detection of the watermarks is also possible without any need for the original cover-image. In addition, the system applies asymmetric cryptographic protocols for different purposes, namely embedding/detecting the watermark and transferring watermarked data. The public key technique is applied for the construction of a one-way watermark embedding and the verification function to identify and prove the uniqueness of the watermark. Legal dispute resolution is supported for the multiple watermarking of a digital image without revealing the confidential keying information.

Robustness and security of digital watermarks

“…If a transformed representation is used, it may be applied to the entire data (the whole image), or on a block-by-block basis. For images, watermarks have been applied to pixels [2], Fourier components [24], whole-image and block-based discrete cosine components [8,23,39], wavelet components [23], and Fourier-Mellin components [25,13]. One use of a transformed representation is to make the components of the data more independent; pixels in an image, for example, are highly correlated locally, which is not true of discrete cosine components.…”

“…For typical data, the largest magnitude components are mostly the low-frequency ones, and marking low frequencies may serve just as well. A possibly better idea for a no-original scheme is to mark mid-frequency components [13,37,19] with the hope that these components are perceptually significant but relatively low-energy. Herigal, et al [13] for example, mark in the Fourier domain.…”

“…A possibly better idea for a no-original scheme is to mark mid-frequency components [13,37,19] with the hope that these components are perceptually significant but relatively low-energy. Herigal, et al [13] for example, mark in the Fourier domain. They avoid "the largest (high energy) components (at about the lowest 10% of the frequencies)" and use "components at the medium frequencies (about next 30%)."…”

“…For example, one may choose to represent a color image using a luminosity-chrominance basis, and mark only the luminosity components, thereby rendering the watermark robust to a color-to-grayscale transformation. A similar idea discussed more fully in the next section is to mark certain Fourier-Mellin components of an image [25,13], thereby obtaining some robustness to the geometric transformations of scaling, cropping and rotation. To enhance security, one may choose to mark only a random subset of the set of components selected for robustness.…”

“…This idea has been refined by Herigal, et al [13]. They first take the logarithm of the luminance levels (to match the human visual system) and then do a Fourier transform.…”