Abstract: A blind image watermarking algorithm based on the multiresolution sub-band decomposition-fast independent component analysis (MSD-FICA) is proposed. In this scheme, the watermark bits are embedded in the middle-frequency components and the low-frequency components simultaneously. The middle-frequency components are selected by adopting the sorting scheme among the high-frequency components. In watermarking extraction, primary component analysis (PCA) is applied for preprocess, then the watermark is extracted by FICA. The experimental results show that the proposed algorithm is robust against JPEG compression, Gaussian noise, and median filtering. The comparison analysis demonstrates that our scheme has better performance than the QIM watermarking scheme.