A new automatic identification technique for OB associations in unresolved galaxies.

We present a new automatic technique based on Principal Component Analysis and Cluster Analysis, with the aim of its application to the identification of those clumps in unresolved galaxies which likely represent regions of star formation. We test the method by applying it to the galaxy M 31, for which there are already several sets of identifications of OB associations based on multi-colour images of resolved stars. We use small-scale digital images of M 31 and compare the associations that we detect from these unresolved data with previously-published large-scale data, finding a rather good agreement. We obtain a strict agreement of our identification with the most compact associations of the original van den Bergh (1964) identification. We then apply the technique to CCD images of the more distant spiral galaxy NGC 2903 and identify 68 OB association candidates.