Digital watermarking of spectral images using SVD in PCA-transform domain

We propose a watermarking method for spectral images. First, the spectral domain of the image is performed with the principal component analysis (PCA) transform and then singular value decomposition is applied to eigenimages from PCA transform. The watermark is embedded by modifying the singular value of one eigenimage. After modification the image is reconstructed, thus containing watermark. We provide analysis of watermark´s robustness against attacks, the attacks include lossy image compression by 3d-spiht and mean filtering. The experiments indicate, that the robustness against attacks is good.