Biomedical Engineering Reference
In-Depth Information
the GGVF snake was specified in the geometric model, when the vector field
is tangent to the snake contour. In such cases there would be no force to push
or pull it in the perpendicular direction (to the vectors). This effect is shown in
Fig. 10.11 (right).
10.4 Region-Aided Geometric Snake
We now describe a novel approach to make the geometric snake much more
tolerant toward weak edges and image noise. It comprises the integration of
gradient flow forces with diffused region forces in the image, resulting in the
region-aided geometric snake:
The gradient flow forces supplant the snake with local object boundary
information. They play a main role in all active contours
4
.
The region forces are based on the global image features and supplant the
snake with global image information.
We show that this combination of forces not only improves the performance of
the geometric snake toward weak edges, but also makes it more immune to noise.
The PDE thus obtained evolves an initial contour toward final convergence under
the influence of both internal forces and boundary-regional image forces, and is
implemented via level sets.
The proposed region force can be generated from any image segmentation
technique. This means that while RAGS is independent of any particular seg-
mentation technique, it is dependent on the quality of the regions produced.
However, we show a good degree of tolerance to (reasonable) segmentation
quality, and that our snake indeed acts as a refinement of the results of the
initial region segmentation. Later in Section 10.7, we introduce the mean shift
segmentation technique presented by Comaniciu
et al
. in [12, 13] which is a
very elegant method to generate region maps for this work. Results will be
presented based on region maps obtained from both the under-segmentation
and over-segmentation options of the software from Comaniciu and Meer's
study.
4
There are notable exceptions to this, e.g. [11].
