A blind wavelet-based watermarking with detail-subband coefficients prediction

Abstract

Recently, the wavelet transform is widely used in multimedia signal processing applications. To provide security solution, the digital watermarking is involved. This study presents a blind wavelet-based watermarking which cooperates with the Human Visual System (HVS) embedding watermarks into detail-subband coefficients. Since the imperceptibility is the most significant issue in watermarking, the approximate band is maintained unchanged, while the other detail subbands are modified to carry information. The perceptual embedded weights for all subbands are determined based on the Just Noticeable Distortion (JND) criterion. The strength of the modification is investigated to provide a compromised result between robustness and image quality. In the decoder, the Least-Mean-Square (LMS) is involved to predict the original detail-subband coefficients and then extract the embedded watermarks. As documented in experimental results, the proposed method provides good robustness and excellent image quality.