Principal Components Data Fusion of Infrared Telescope Multi-Spectral Images of the Galactic Center

摘要

As contrasted with monochromatic images, multi-spectral images contain a wealth of information helpful to phenomenological interpretation by scientists. Image data fusion techniques are useful to uncover the underlying correlations in the complex data that, in turn, facilitates image enhancement, feature recognition, and classification. We report in this paper the results of a Principal Components Analysis (PCA) based upon covariance and correlation matrices between astronomical images taken in four infrared spectral passbands. We processed the image data set using unstandardized PCA (covariance) and the standardized PCA (correlation). The data set was obtained from the spatial imaging radiometer, SPIRIT III, flown in 1996 aboard the Midcourse Space Experiment (MSX) satellite. The Galactic Center was imaged in four mid-infrared passbands centered on wavelengths of 8.28, 12.13, 14.65, and 21.34 μm. Both PCA approaches yielded viable information in all four principal components, including the high-order components. The first principal component demonstrated the success of PCA as an image fusion technique. Higher order principal components provided feature discrimination between stars and star-like dust concentrations of different spectral classification, and of nebulae and molecular clouds with varying physical and chemical properties.