Hyperspectral image fusion using 2-D principal component analysis

Abstract

In this work, a novel fusion scheme for efficient representation of a hyperspectral dataset in an informative color image is proposed using 2D-PCA. The fusion approach is based on partitioning the hyperspectral dataset into subgroups of bands, and image covariance matrix is directly applied on the 2D matrices of each spectral band. The resulting image representation offers the ability to effectively discriminate information, providing advanced performance in terms of multiband representation. Experimental results, provided on two hyperspectral dataset acquired by CHRIS sensor and the AVIRIS instrument, demonstrate the advantage of the proposed work.