Watermarking through color image bands decorrelation
Alessandro Piva, F. Bartolini, Luca Boccardi, V. Cappellini, Alessia De Rosa, Mauro Barni · 2002
Image watermarking has received great attention. Nevertheless, researchers are still faced with several problems: the extension of gray-level watermarking to the color case is one of them. To get rid of the problems raised by the correlation among image color bands, a new approach based on the decorrelation property of the Karhunen-Loeve transform (KLT) is proposed. The KLT is first applied to the RGB components of the image, then watermarking is performed independently in the DFT domain of the KL-transformed bands. Watermark embedding is achieved by modifying the magnitude of mid-frequency DFT coefficients according to an additive-multiplicative rule. On detection, KL decorrelation is exploited to optimally detect the watermark presence by relying on Bayes statistical decision theory.