An image database III. Automatic extraction for millions of galaxies

This pager presents a method for extracting a catalogue of galaxy candidate s from the Digitized Sky Survey (DSS). The method is based on a functional analysis applied on each individual plate. The standard deviation of pixel optical densities versus the inverse of surface area leads to a diagram in which extended and star-like objects are well separated. This diagram is us ed for a preliminary recognition. Then, a filtering process is applied usin g a Neural Network method associated with a training sample built with well identified objects. The main catalogue gives coordinates, total magnitude, isophotal diameter, axis ratio, position angle for 2 772 061 galaxy candid ates. The method favors the detection of normal galaxies. This creates a bi as against compact high surface brightness galaxies.