Back to Volume
Paper: Spectral and Spatial Bayesian Segmentation of Hyperspectral Astronomical Data Cube
Volume: 371, Statistical Challenges in Modern Astronomy IV
Page: 409
Authors: Petremand, M.; Collet, Ch.; Flitti, F.
Abstract: Nowadays the enhanced spectral resolution provides access to large hyperspectral cubes. This high spectral resolution allows the astronomers to highlight physical and chemical properties in order to characterize objects through the cube analysis. Each pixel within the hyperspectral image is then described thanks to its spectral behavior. The segmentation of these hyperspectral images is carried out in order to obtain a single map highlighting spectral features by using a set of spectral signatures, associated with a Markovian model to regularize the classification process. We propose a new method, that has been established on the Mean-Shift-based approach coupled with a Markovian modeling, to reduce, segment and finally extract the main spectral behaviors within a hyperspectral data cube. It has been tested here on a simulation of a hyperspectral galaxy field. This kind of tool will be of great interest in the next decades to process new hyperspectral data with spatial resolution.
Back to Volume