Watermarking Algorithm Based on Curvelet Transform and Human Visual Model

Fast Discrete Curvelet Transform (FDCT) is a multiscale, multidirection and shift-invariant over complete transform. The paper proposes a blind watermarking algorithm in the FDCT domain. A simple but effective human visual model in FDCT domain is described. By modelling watermarking as a communication process. The experimentations show that this domain has higher capacity than the wavelet (include complex wavelet and real wavelet) and spatial domains. FDCT has better direction selectivity and less redundancy than wavelets, and has shift-invariant property. Watermarking algorithm based on curvelet transform improves the robustness of watermark at gaussian white noise attack. Hence that curvelet provide a potentially good domain for watermarking.