Pitiot Alain, Delingette Hervé, Toga Arthur W, Thompson Paul M
Epidaure, INRIA, 2004 route des lucioles BP 93, 06 902 Sophia-Antipolis, France.
Inf Process Med Imaging. 2003 Jul;18:25-37. doi: 10.1007/978-3-540-45087-0_3.
We propose a learning method which introduces explicit knowledge to the object correspondence problem. Our approach uses an a priori learning set to compute a dense correspondence field between two objects, where the characteristics of the field bear close resemblance to those in the learning set. We introduce a new local shape measure we call the "observed transport measure", whose properties make it particularly amenable to the matching problem. From the values of our measure obtained at every point of the objects to be matched, we compute a distance matrix which embeds the correspondence problem in a highly expressive and redundant construct and facilitates its manipulation. We present two learning strategies that rely on the distance matrix and discuss their applications to the matching of a variety of 1-D, 2-D and 3-D objects, including the corpus callosum and ventricular surfaces.
