Multiple Digital Watermarking Scheme Based on ICA

This paper proposes a novel approach to multiple digital watermarking of images based on independent component analysis (ICA). Firstly, discrete wavelet transform (DWT) was applied to one sub-image extracted from the original image orderly, at the same time, multiple independent watermark signals mixed with ICA and generated mixed signals. Then a signal was chosen from the mixed signals as watermarking embedding into the intermediate frequency by mixing with ICA. At extracting stage, using the same rule as embedding stage, extracting the sub-image from watermarked image, then every embedded watermark signal can be extracted through twice demixed with ICA by inverse embedding process. Using the watermarking standard testing tool Stirmark Benchmark 4.0 to attack the watermarked images. The simulation results show that multiple watermarks are extracted perfectly, and also demonstrate robustness and invisibility of the method.