|Limits of Transient Detection in Mixed-Resolution Image Analysis|
Darren Homrighausen, Colorado State University
Monday, September 10, 2012
4:00pm, room 223 Weber Building
In this talk, we consider the problem of detecting transients in astronomical images. To do the detection, two images of the same object, or scene, are compared to find differences. The idea is that intrinsic differences between the images represent transient phenomena of interest. Transients, such as planets or supernova, can have many different forms -- some unknown -- and detection is complicated because the two images being compared usually differ substantially in quality. Typically, the ``science image'' has poorer resolution, more blurring, and more noise than the ``reference'' image. Disentangling spurious differences due to quality variations from differences caused by transients requires a transformation of the images to make them comparable. We develop a method, FASTDetect, that automatically identifies transients by estimating the transformation between science and reference images. We analyze data from the Sloan Digital Sky Survey to show the method's effectiveness. Also, we derive theoretical conditions characterizing under which conditions specific transients can be detected.