The European Southern Observatory has catalogues with image data available from http://www.eso.org/qi/, you will have to register before you are able to access them.
I'd suggest you look at other observatory's websites for their data. You will have to look past the pages targeted at the general public and find links for data or science, or user portal or something like that. They are sometimes difficult to find.
Ideally you would have a standard annotated data set of images for classification so that you will be able to compare your results with others. Unfortunately I'm not aware of any standard sets.
For literature on using pattern recognition on astronomical images, http://astrometry.net/biblio.html is a good resource. They've created an ML system that is able not only to distinguish stars from other object classes but also able to identify which stars are visible in an image! Very interesting research.
NB For classifying stars in clusters, the problem is likely not classification, but more segmentation.