Redundant watermarking using wavelet packets

This paper presents a new blind logo watermarking algorithm based on wavelet packets. For increasing robustness and security together, the logo watermark is scrambled n times and each time with a different private key to generate n scrambled copies of the logo watermark. To embed the logo watermarks, the host image is decomposed by wavelet packets transform and then the scrambled logos are embedded in selected sub-bands of wavelet transform by quantization technique. Knowing the quantization step-size, the private keys, position of the selected sub-bands and dimensions of the logo watermark, the watermark is extracted without any need to have access to the original image. The extracted watermarks are combined to afford a new perfect version of the watermark. Robustness of the proposed algorithm was tested against the following attacks: lossy compression, noise addition, filtering, quantization, cropping, rotating, and scaling. The promising experimental results are reported and discussed.