[Insight-users] Kd-tree implementation in ITK is buggy and unreliable: Bucket Size

Luis Ibanez luis.ibanez at kitware.com
Fri Apr 25 11:17:55 EDT 2008


Looking more into the implementation of the KdTree,...
I'll take back the statements regarding the bucket sie parameter.

In fact, the KdTree and KdTreeGenerator should work properly
regardless of the seetting in this parameter.

What large bucket sizes are actually achieving is forcing the
KdTree to have a single node at the root with all the measurement
vectors stored there.  Which, of course, defeats the purpose of
the KdTree as a structure for accelerating the search for close
points.

We have observed that if itkKdTreeTest2 is run with bucket sizes
smaller than 3 it produces segmentation faults.

.... still tracking the problem....


    Luis




More information about the Insight-users mailing list