Biomedical Engineering Reference
In-Depth Information
(a)
(b)
(c)
(d)
Figure 3. Noise removal and segmentation via moving under curvature: (a) image with
mild noise; (b) after removing the noise; (c) image with strong noise; (d) after removing
the noise.
the CFL condition is necessarily unstable, but satisfying the CFL condition does
not necessarily guarantee stability.
In the cases where the gradient information is strong enough, the fast marching
method can be directly used for segmentation. An extra stopping criterion has to
be defined. Awidely used definition of the gradient-based speed is e −α|∇I| , where
α is a constant. A simple threshold value could be defined on the gradient. The
speed factor could also be defined based on intensity itself. When the front touches
the pixel with a gradient value above the threshold, the front point is frozen. If all
the candidate pixels on the front are frozen, the Fast Marching Methods stop.
In the following, several segmentation examples using this method are illustrated.
Search WWH ::




Custom Search