Tessellated surface reconstruction from 2D contours

Abstract

This paper presents a new triangulation method to define surfaces of 3D object from parallel 2D contours. These contours represent the boundary of a human organ segmented from 2D images acquired from radiological volumetric data using ultrasound (US), computer topography (CT) and magnetic resonance (MR) imaging. Many papers have identified and looked into the problems of generating surfaces from 2D contours. They are correspondence, tiling, branching and surface-fitting problems. Our new algorithm dealt with these problems in three steps. First, several adjacent contour mapping positions are used to reveal the correspondence of contours. Second, a tessellation algorithm approximates contours into line segments. Finally, surface meshes, in the form of strips and fans, use the gradient of line segments to optimise surface quality without affecting the rendering time.