Related Experiment Videos
A vision-based masking model for spread-spectrum image watermarking
Martin Kutter1, Stefan Winkler
1Signal Processing Laboratory, Swiss Federal Institute of Technology, 1015 Lausanne, Switzerland. martin.kutter@alpvision.com
Summary
This study introduces a perceptual model for robust image watermarking. Optimal watermark detection performance is achieved at moderate embedding densities, not maximum.
Area of Science:
- Digital image processing
- Computer vision
- Human visual perception
Background:
- Digital watermarking is crucial for copyright protection and data integrity.
- Existing methods often struggle to balance robustness with visual quality.
- The human visual system (HVS) plays a key role in watermark imperceptibility.
Purpose of the Study:
- To develop a perceptual model for embedding spread-spectrum watermarks in images.
- To optimize watermark robustness and visual quality by considering HVS characteristics.
- To investigate the impact of watermark embedding density on detection performance.
Main Methods:
- A perceptual model incorporating local isotropic contrast and a masking model was developed.
- Spread-spectrum watermarks of variable amplitude and density were embedded.
- Watermarking was performed on luminance and blue channels of color images.
- Robustness was evaluated against varying embedding densities.
Main Results:
- The proposed model enables the insertion of more robust watermarks while maintaining visual quality.
- Embedding watermarks in the blue channel of color images was compared to luminance.
- Maximum watermark density did not consistently yield the best detection performance.
Conclusions:
- The perceptual model effectively balances watermark robustness and imperceptibility.
- Moderate watermark embedding densities are preferable for optimal detection performance.
- The findings offer insights for designing more effective digital image watermarking systems.
