ABSTRACT
Isosurfaces have long been used as a meaningful way to represent feature boundaries in volumetric datasets. Surface representations of volumetric datasets facilitate visual comprehension, computational analysis, and manipulation. The earliest beginnings in this field can be traced to microscopic examination of tissues. For viewing opaque specimen, it was considered desirable to slice the sample at regular intervals and to view each slice under a microscope. The outlines of key structures in such cross-sectional slices were traced on to transparent sheets that were then sequentially stacked with transparent spacers in between them [WC71, LW72]. A user holding such a translucent stack would then be able to mentally reconstruct the location of the surface interpolating the contours on successive slices. It is, therefore, only natural that one of the first algorithms for surface reconstruction from volumes proceeded by interpolation between planar contours [FKU77]. Fuchs et al.’s method proceeded by casting the problem of optimal surface reconstruction between planar contours as one of finding minimum cost cycles in a directed toroidal graph. Their method computed the minimum-area interpolating surface between the given planar contours. The advent of x-ray computed tomography (CT) in 1973 through theoretical advances
#2 ✐
✐
✐
✐
✐
✐
by Cormack [Cormack63,Cormack64] and their practical realization by Hounsfield [Hounsfield73] made it significantly easier to acquire highly detailed cross-sectional images. The use of magnetic fields and radio waves to create images of physical objects by Lauterbur [Lauterbur73] and Mansfield [GGM74] led to the development of magnetic resonance imaging (MRI). The obvious implications of such volumetric imaging technologies in medical diagnosis and therapy sparked a lively interest among radiologists and scientific visualization researchers and greatly accelerated research in a variety of volume visualization and analysis tools and techniques [Levoy88, Kaufman91]. The Visible Human Project (VHP) of the National Library of Medicine (NLM) has greatly spurred the development of fast volume visualization techniques, including fast techniques for isosurface extraction by making available 1971 digital anatomical images (15 GB) for the VHP male dataset and 5189 digital images (39 GB) for the VHP female dataset [Yoo04]. Some isosurfaces extracted from the VHP male dataset appear in the following text [Lorensen01] (Figure 27.1).
