2.2 Skeleton Pruning
After segmenting a flower from background region, we then extract a skeleton of a flower for effective
representation. For generating the skeleton of a flower, we have adopted the work proposed by Bai et.al, [10].
Initially, we binarize the gray scale flower image and then by applying some morphological operations, we will get
the skeleton of a flower image. Then, the Discrete Curve Evolution (DCE) is used to simplify the polygons obtained
from un-pruned flower skeleton image. It recursively removes the least relevant polygon vertices, where the
relevance measure is computed with respect to the actual partially simplified versions of the polygon. Then the
skeleton is pruned so that only branches ending at the convex DCE vertices will remain. The pruned skeleton is
guaranteed to preserve the topological relationship of the shape and it is robust to noise and boundary deformation.
The main benefit of using DCE is the fact that DCE is context sensitive.
After obtaining a skeleton from a flower, we extract some points from skeleton viz., end points and junction
points that are useful in generating Delaunay triangulation. A skeleton point having only one adjacent point is called
an endpoint (the skeleton endpoint); a skeleton point having three or more adjacent points is called a junction point.
The figure 1 shows the detected endpoints and junction points for a given flower skeleton. In the subsequent section
we can see how these points are helpful in generating Delaunay triangulation.
2.2 Skeleton Pruning After segmenting a flower from background region, we then extract a skeleton of a flower for effective representation. For generating the skeleton of a flower, we have adopted the work proposed by Bai et.al, [10]. Initially, we binarize the gray scale flower image and then by applying some morphological operations, we will get the skeleton of a flower image. Then, the Discrete Curve Evolution (DCE) is used to simplify the polygons obtained from un-pruned flower skeleton image. It recursively removes the least relevant polygon vertices, where the relevance measure is computed with respect to the actual partially simplified versions of the polygon. Then the skeleton is pruned so that only branches ending at the convex DCE vertices will remain. The pruned skeleton is guaranteed to preserve the topological relationship of the shape and it is robust to noise and boundary deformation. The main benefit of using DCE is the fact that DCE is context sensitive. After obtaining a skeleton from a flower, we extract some points from skeleton viz., end points and junction points that are useful in generating Delaunay triangulation. A skeleton point having only one adjacent point is called an endpoint (the skeleton endpoint); a skeleton point having three or more adjacent points is called a junction point. The figure 1 shows the detected endpoints and junction points for a given flower skeleton. In the subsequent section we can see how these points are helpful in generating Delaunay triangulation.
การแปล กรุณารอสักครู่..
