This work addresses a galactic morphology problem from
astronomy: given a noisy, blurred image of a galaxy, how
can we determine its structure automatically and quickly?
Solving this problem is of importance to astronomy, since
the formation of galaxy structure is poorly understood, and
there are millions of such images unanalyzed due to
I will describe a new algorithm which performs the search for fits in eigenspace, consisting of a basis of 10 or so "eigengalaxies." The eigenspace is pre-populated with many synthetic galaxies, and the fits are to the nearest neighbors of the image in this space. Local PCA is used to deal with the blurring effects of the atmosphere as well as small x/y offsets in the image. We achieve significant improvements in speed over previously published techniques.
Back to the Main Page
Charles Rosenberg Last modified: Fri Sep 12 17:08:40 EDT 2003