Biomedical Engineering Reference
In-Depth Information
Chapter 7
Improving the Initialization, Convergence, and
Memory Utilization for Deformable Models
Gilson A. Giraldi
1
, Paulo S. Rodrigues
1
, Leandro S. Marturelli
1
, and
Rodrigo L. S. Silva
1
7.1 Introduction
In this chapter our aim is twofold. Firstly, we point out some limitations of
deformable models for medical images and analyze recent works to overcome
these limitations. Next, we offer new perspectives in the area, which are part of
our current research in this field.
Deformable models, which include the popular
snake models
[42] and de-
formable surfaces [19, 48], are well-known techniques for tracking, boundary
extraction, and segmentation in 2D/3D images.
Basically, these models can be classified into three categories: parametric,
geodesic snakes, and implicit models. The relationships between these models
have been demonstrated in several works in the literature [57, 75].
Parametric deformable models consist of a curve (or surface) which can
dynamically conform to object shapes in response to internal (elastic) forces
and external forces (image and constraint ones) [6].
For geodesic snakes, the key idea is to construct the evolution of a con-
tour as a geodesic computation. A special metric is proposed (based on the
gradient of the image field) to let the minimal length curve correspond to the
desired boundary. This approach allows one to address the parameterization
1
National Laboratory for Scientific Computing, Brazil
359
