Reduction and Segmentation of Hyperspectral Data Cubes

摘要

The reduction and segmentation of multiwavelength images become problematic when the number of bands increases. Integral field spectroscopy and other instrument designs allowing for enhanced spectral and spatial resolution lead to extremely large hyperspectral data cubes (typically 370 million pixels per exposure for the MUSE instrument). New analysis tools jointly exploring spectral and spatial features are required. We propose a new approach, based on the Mean-Shift method (Comaniciu & Meer 2002), to reduce the dimensionality of large data cubes and extract the main spectral patterns. A set of spectra extracted from the cube is used as an initial reference basis. Each spectrum in the observation is projected on this basis, and represented by a vector of projection coefficients or weights. The Mean-Shift method is then carried out for the whole dataset to find the modes in the projection space. These modes are selected for a new projection basis and the algorithm is iterated until convergence. The distance between two spectra is defined as the angle between their related vector coefficients to increase efficiency: this speeds up the convergence and gives more weight to the comparison of spectral patterns, minimizing the effect of the average intensity of each spectrum. This approach has been tested on simulated data. It is promising, especially for very high spectral resolution data cubes.