1999ApJ...511..242G

Our study aims at the characterization of H I gas cloud images for classification purposes using their brightness temperature distribution. The morphology and velocity structure of the gas cloud is not taken into account in this first study. The analysis is based on spatial contiguity at a chosen radial velocity value. It is assumed that the form brightness temperature distribution of each cloud is dependent on the physical process that created it. We wish to identify the existing relations between different clouds as a first step toward understanding the evolutionary and physical processes that form them. Our study starts with an exploratory analysis based on principal components analysis and on cluster analysis. Because the inherent signal in each image is noise, the characterization is then studied by using a method of density estimation by deconvolution. The clustering of the images is done by considering single linkage and complete linkage criteria of cluster aggregation.