RDWT/ICA for image authentication

This paper proposes a new approach to watermarking multimedia products by redundant wavelet transform (RDWT) and independent component analysis (ICA). For watermark security, embedded logo watermarks are encrypted to random noise signal. To embed logo watermarks, the original image is decomposed by RDWT, and watermarks are embedded into middle frequency subbands. The perceptual model is applied with a stochastic multi-resolution model for adaptive watermark embedding. This is based on computation of a noise visibility function (NVF) which has local image properties. We also propose an intelligent ICA-based detector which directly extracts watermarks in spatial domain. A novel characteristic of this detection is that it does not require the transformation process to extract the watermark. The experimental results show that logo watermarks are extracted perfectly, and also demonstrate the robustness of the method