Spectral/Spatial Hyperspectral Image Compression
Bharath Ramakrishna, Antonio Plaza, Chein-I Chang, Hsuan Ren, Qian Du, Chein-Chi Chang · Kluwer Academic Publishers eBooks · 2006
This chapter investigates the applicability of direct application of 3D compression techniques to hyperspectral imagery and develops PCA-based spectral/spatial compression techniques in conjunction with the virtual dimensionality (VD) for hyperspectral image compression where the VD is used to estimate number of principal components required to be preserved. In particular, we conduct computer simulations based on a synthetic image and real image experiments to demonstrate that simple PCA-based spectral/spatial lossy compression techniques can perform at least as well as 3D lossy compression techniques in applications such as mixed pixel classification and quantification. This interesting finding provides evidence that PCA-based spectral/spatial compression can be as competitive as the 3D compression for hyperspectral image compression. Additionally, this chapter also further demonstrates that the number of PCs required to be preserved by lossy compression is crucial and the proposed VD provides a much better estimate than the commonly used criterion determined by the sum of largest eigenvalues. For more details we refer to [31].