In this paper, we propose a method of embedding the spectral information into an RGB image by robust watermarking. The difference between an original image and a multispectral image estimated from the RGB image by Wiener estimation is referred to as the residual. The estimated spectral residual data are compressed by JPEG2000 and embedded into the wavelet coefficients of chrominance channel. When a multispectral image is reconstructed from an RGB image, our proposed method can improve the quality of the reconstructed multispectral image in comparison with a multispectral image estimated by Wiener estimation because the residual data can be extracted from the watermarked RGB image. Additionally, if the watermarked RGB image is compressed with a lossy compression method, the proposed method can obtain the high-quality reconstructed multispectral image because this method uses robust watermarking. The experimental results show that there is almost no significant difference in the RGB image.
Kazushige Banzawa, Kazuma Shinoda, Madoka Hasegawa, "Robust Multispectral Data Hiding in RGB Image Using Digital Watermarking" in Proc. IS&T 24th Color and Imaging Conf. , 2016, https://doi.org/10.2352/ISSN.2169-2629.2017.32.213