A new methodology to automatically segment biomedical images

Abstract

This paper presents a new approach for 2D object segmentations using an automatic method applied on images with severe noise conditions and locating objects with a very high degree of deformation. We use a physically-based shape model to obtain a deformable template, which is denned on a canonical ortogonal coordinate system. The proposed methodology works from a set of samples and from the output of an edge detector to segment the objects by using a reformulated Hough transform (automatic initialization) together with an optimization procedure (on a learned surface of deformation). Results from biomedical images are presented.