Bowyer/Watson algorithm: continue
by Yi Zhang
Here is Delaunay triangulation for one thousand random points, using Bowyer/Watson algorithm in my last post.
The second plot above comes up when the enlosing trianlge is erased from the first one, which usually leaves some concave spaces at the boundary, instead of convex hull shown in the third plot. It seems that the error of those random concaves diminshes as the number of random points arises, I am not sure whether one can use this fact to ignore those concaves in finite element method.