Efficient implementation of KL entropy estimator for image registration

                         Ivan Vnucko

Fundamental problem of mutual information image registration is the
estimation of entropy and mutual information. We reimplemented the
Kozatchenko-Leonenko estimator based on nearest neighbor
distances. All-nearest neighbor search uses a kD tree with Best Bin
First node traversal. Our main goal was the speed. In the presentation
we will shortly introduce the KL estimator and the BBF kD tree,
describe the implementation and results of experiments.