Biomedical Engineering Reference
In-Depth Information
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
Figure 4.
Segmentation of a square pattern from background. The first row shows initial-
izations for both algorithms. The second row shows the results of geodesic active regions
[11, 12, 13]. The third row shows results of the proposed algorithm. (a-c): Gaussian noise,
background mean/variance = 10/100, square mean/variance = 10/225. (d-f ): Gaussian
noise, background mean/variance = 0/0.1, square mean/variance = 0/1 (the curve in (e) will
not stop and expands out of the image domain). (g-i): salt and pepper noise, background
and square mean = 1 noise density = 0.5.
segmentation for scalped brain MR images. Figure 7 shows the 3D segmentation
results. The zero level set of two level set functions are both initialized as multiple
spheres. In Figure 7a, 27 small spheres (gray) are initialized for level set function
φ
0
, four larger (deep gray) spheres are initialized for level set function
φ
1
. Both
level set functions are updated according to Eq. (27)-(28). The final results are
