Tensor Decomposition and PCA Jointed Algorithm for Hyperspectral Image Denoising

Denoising is a critical preprocessing step for hyperspectral image (HSI) classification and detection. Traditional methods usually convert high-dimensional HSI data to 2-D data and process them separately. Consequently, the inherent structured high-dimensional information in the original observations may be discarded. To overcome this disadvantage, this letter tackles an HSI denoising by jointly exploiting Tucker decomposition and principal component analysis (PCA). A truncated Tucker decomposition method based on noise power ratio (NPR) analysis and jointed with PCA is presented. We call this jointed method as NPR-Tucker+PCA. Experimental results show that the proposed method outperforms existing methods in the sense of peak signal-to-noise ratio performance.