Image Processing Reference
In-Depth Information
the plane in Figure 6(a) serves only as a mere illustration of the principle used by the tech-
nique, and do not correspond to the one used to obtain the classification in Figure 6(b) .
FIGURE 6 (a) Planar segmentation by RANSAC with a threshold of ± 5 mm and (b) dental
model segmented.
This method finds the highest concentration of points in a planar model based in a planar
hypothesis. In this case the highest concentration of points is found on the teeth's surface.
4.2 Gum Extraction Using Region Growing
The results obtained by applying the algorithm of region growing are shown in Figure 7 . The
stop criteria are given comparing the points normal and then the curvature of the points. The
input parameters are shown in Table 1 .
FIGURE 7 (a) Example of curvature in point clouds of dental models, (b) 3D dental model
segmented using a Region Growing segmentation.
Table 1
Input Parameters for Region Growing Segmentation
c th θ th N Ω {.} P
Tree Point cloud
In order to improve understanding and visualization, the point cloud of dental model is
drawn with colors representing their curvature Figure 7(a) and the result obtained applying
the Region Growing segmentation technique in a 3D dental model is illustrated in Figure 7(b) .
For further information regarding neighbor search process and Kd-trees please refer to PCL,
API Documentation [ 21 ] or the Flann library documentation [ 22 ] .
Figure 7(b) shows some points colored in aqua blue (light gray in the print version), which
are not targeted because of their high curvature or because the region generated by the region
growing process do not contain enough data points to be considered as such.
Clearly this method requires some refinement that would allow merging some regions.
However, this is an acceptable approximation which can be used instead of usable as start
point for further refinement.
Search WWH ::

Custom Search