Hyperspectral image cubes are usually obtained from direct measurement (e.g. satellites) and by using numerical reconstructions from imaging spectrometers. Such objects are three dimensional and consist of 2D spatial data layered by wavelength in the third dimension. The dominant image processing tasks for such information are compression and feature recognition. These tasks go hand-in-hand as most often these objects contain a huge amount of information that need to be processed further (and often very quickly) corresponding to the application. Wavelet transforms which have been utilized successfully for signals and images are applied here. The higher number of dimensions furnishes a number of different ways to do these transforms and some of these ways are more natural for hyperspectral data processing.

© 2003 Optical Society of America

PDF Article
More Like This
Design and Characterization of an Optical Chip for Data Compression based on Haar Wavelet Transform

Cátia Pinho, Ana Tavares, Guilherme Cabral, Tiago Morgado, Ali Shahpari, Mário Lima, and António Teixeira
Th2A.9 Optical Fiber Communication Conference (OFC) 2017

Compression of digital holograms using 1-level wavelet transforms, thresholding and quantization of wavelet coefficients

P A Cheremkhin and E A Kurbatova
W2A.38 Digital Holography and Three-Dimensional Imaging (DH) 2017

Feature extraction using wavelets to improve the estimation of material fractional abundances

Kris Hamada, S. Maile Giffin, and Charles L. Matson
FTuB5 Frontiers in Optics (FiO) 2004