EXTRACTION OF WATERMARKS BASED ON INDEPENDENT COMPONENT ANALYSIS

We propose a new logo watermark scheme for digital images which embed a watermark by modifying middle-frequency sub-bands of wavelet transform. Independent component analysis (ICA) is introduced to authenticate and copyright protect multimedia products by extracting the watermark. To exploit the Human visual system (HVS) and the robustness, a perceptual model is applied with a stochastic approach based on noise visibility function (NVF) for adaptive watermarking algorithm. Experimental results demonstrated that the watermark is perfectly extracted by ICA technique with excellent invisibility, robust against various image and digital processing operators, and almost all compression algorithms such as Jpeg, jpeg 2000, SPIHT, EZW, and principal components analysis (PCA) based compression.