Robust Data Hiding With Image Invariants

January 22, 2007 -- TRLabs has developed a new technique for digital watermarking that has been shown to be more robust against geometric distortions.

Benefits
• Hundreds of bits can be embedded imperceptibly.
• Embedded data is robust to geometric attacks as well as other common distortions.
Background
Data hiding, or digital watermarking, is a technique of embedding some additional information into a host signal such as an image, an audio or video clip, by imperceptibly modifying some features of the host signal. It has been regarded as a promising means to address such issues as authorship identification, content authentication and illegal copy tracing etc. Amongst various requirements of a watermarking system, the watermark robustness is crucial in many situations. A watermark robust to geometric distortions as well as other processing such as lossy compression is particularly difficult to design, and has not been well addressed so far.

Using invariant image features is an effective approach to the watermark robustness against geometric distortions, which has been the subject of research in recent years. Some progress has been made in this direction, such as the Fourier-Mellin domain based approach and the moment based approach. More recently, Zernike moment (ZM) was proposed for invariance watermarking. However, with the conventional Cartesian method for ZM computation, the invariance property of ZMs is far from ideal.

TRLabs has focused on the investigation of embedding a multi-bit watermark in the magnitudes of ZMs, in an effort to combat geometric manipulations of images, in particular, rotation, flipping, and scaling. TRLabs proposes a polar coordinates based approach to accurate ZM computation, and hence the improvement of the invariance property. Due to the improved invariance of ZM magnitudes, TRLabs has achieved better watermark robustness against image rotation, flipping, and scaling, as well as a number of other common manipulations such as lossy compression and low-pass filtering.
Project Status
TRLabs has developed a novel method of calculating Zernike moments that has been shown to greatly improve the robustness of watermarks embedded in images.
Future Developments
In addition to image rotation, scaling, translation and flipping, the watermark robustness to other attacks such as image cropping and general affine transformations will be examined. An extension of the watermark to video and audio objects will also be considered.
Research Supervisors
Mirek Pawlak t: 204.488.5605 e:
For more information on this, and other, TRLabs initiatives please contact
Dr. Vinod Ratti
7th Floor, 9107 - 116 Street
Edmonton, AB, Canada T6G 2V4
t: 780.441.3812 f: 780.441.3600 e:
Homepage: http://www.trlabs.ca/trlabs/technology/

Emerging Technology Bulletins are published by TRLabs periodically to inform our sponsors of early stage technologies and invite collaboration towards further development.

Login to see more detailed information on this ETB

See more technology bulletins »