How to recover smooth object boundaries in noisy medical images

Thomas Strohmer, Thomas Binder, and Michael Süssner

Diagnostics in medicine is often based on analysis of medical images. Extraction of the shape of objects is one basic step for further image analysis. It consists of two steps: (i) edge detection and (ii) contour recovery based on the detected edge points. In this paper we focus on the second step. We describe the boundary of an object by trigonometric polynomials and present a fast and robust method for the approximation of the boundary from a given set of nonuniformly distributed noisy edge points. We show how the proposed technique can be efficiently generalized to 3-D surface recovery. Examples from 2-D and 3-D echocardiography are given.

Keywords: algorithms, irregular sampling, numerical work, surface approximation, contour recovery, medical imaging, left ventricle
Published: in IEEE Proc. ICIP'96, 1:331-334, Lausanne, 1996

Download the paper as a GNU-ziped postscript file (167186 bytes).