A modified spatial spread spectrum method for digital image watermarking

Spread spectrum is one of the most effective approaches for digital image watermarking because it satisfies the imperceptiveness, robustness as well as security requirements. Unfortunately, this method may be affected by interference from the cover image and completely fails in the case of having no watermark in the tested image. In addition, some proposed improvements to eliminate this interference are invalid for spatial image watermarking which obtains a low cost and is easy to adapt to real-time applications. This paper proposes a modified spatial spread spectrum method to overcome these disadvantages. By removing the mean value in the tested image, the proposed method reduces the interference and achieves a better error probability. Moreover, by using a flexible threshold in the detector, our method can work successfully even if the tested image is pure. We not only analyze and derive the closed-form expressions for the optimal threshold and the error probability, but also investigate various images and some common image attacks including noise, compression and filtering.