Using deformable models to segment complex structures under geometric constraints

Abstract

In many problems of medical or geophysical interest, when trying to segment an image, one has to deal with data that exhibit very complex structures. This problem occurs when images have discontinuities: in medical imaging (fractures radiography), in geophysics (segmentation of a set of layers and faults) etc. To solve this problem, we present a segmentation method which uses deformable models. The originality of the method is that we have interpolation data and triple points that involves making some geometric constraints on the model. We also propose a method for noise removal because it is well known that most of these images are noisy, that could hinder the segmentation. Numerical results on geophysical images are given.