Image-Feature Based Second Generation Watermarking in Wavelet Domain

An image-feature based second generation watermarking scheme is proposed in this paper. A host image is firstly transformed into wavelet coefficients and features are extracted from the lowest approximation. Then a watermark sequence is inserted in all high frequency coefficients corresponding to the extracted featured approximation coefficients. Original host image is not needed in watermarking detection, but the featured approximation coefficients position is necessary for robust detection. The correlation between the embedded watermark and all high frequency coefficients of a possibly corrupted watermarked image corresponding to the approximate coefficients at the same position as the original featured approximation coefficients is calculated and compared to a predefined threshold to see if the watermark is present. Experimental results show the watermark is very robust to common image processing, lossy compression in particular.