An adaptive image watermarking algorithm leveraging spatial and frequency, domain techniques

摘要

This paper proposes a new adaptive watermarking algorithm leveraging combined knowledge of spatial and frequency domains to ensure the robustness and security of the embedded watermark and the transparency of the carrier image. The scheme makes use of the perceptual characteristics of human visual system and the local correlation of the image. The wavelet coefficients of the image are divided into multiple clusters by constructing the pixel feature vector and applying the fuzzy clustering algorithm. Each cluster is assigned different embedding strength to adapt to each sub-image adaptively embedding watermark. Finally, using the vote principle during extraction, the final extracted watermark is got though the synthesis of the three extracted watermark image. Our experimental results show that the proposed algorithm not only has good transparency, but also has better robustness for image attacks, such as superimposed noise, JPEG compression, median filter and geometric cropping.