A Secure Wavelet based Watermarking Approach using Psycho-Visual Redundancy in Image

The principal requirements of any digital watermarking algorithm are imperceptibility, robustness, security and capacity of the watermark [2], [9]. Imperceptibility and robustness are inherently conflicting in nature. Over numerous trials of digital watermarking of multimedia, it has been found that embedding a watermark with high strength also decreases its obscurity and makes the watermark more prone to detection. On the other hand, a watermark embedded with a low strength improves the imperceptibility but causes degradation of robustness against signal processing, geometric and feature distortive attacks. This paper proposes a secure approach for watermarking using the Image Partitioning approach, which exploits the psycho-visual redundancy in the image to embed the watermark. Using the chaotic scrambling based Arnold Transform, watermark is first scrambled and then embedded in the sub-band of the image obtained using the Discrete Wavelet Transform and characterized by the lowest frequency. This improves the security as well as enhances the imperceptibility of the watermark, without significant deviation in visual appearance of the watermarked image from the cover image. Further, frequency domain embedding leads to better robustness and stability. The method is used to perform watermarking of grayscale images and performs well when subjected to common distortions, geometric, order-statistic filtering and signal processing attacks. This was determined by calculating the Peak Signal to Noise Ratio of the watermarked cover image and the retrieved watermarks and the Normalized Cross-Correlation of the embedded and extracted watermarks.