I work on empirical inference on image data. Taking a photo is a measurement process, which can be affected by different types of errors. This includes motion blur, optical aberrations, and noise.
With prior knowledge of both the typical image content and the source of corruption, the measurement error can be removed to a large extent. In particular, I'm interested in using machine learning techniques to extract and model this prior knowledge.
CJ. Schuler, M. Hirsch, S. Harmeling, B. Schölkopf
(2012). Blind Correction of Optical Aberrations In: Computer Vision - ECCV 2012, LNCS Vol. 7574, (Ed) A Fitzgibbon, S Lazebnik, P Perona, Y Sato, and C Schmid, Springer, Berlin, Germany, 187-200, ISBN: 978-3-642-33711-6, 12th IEEE European Conference on Computer Vision, ECCV 2012