[ImageJ-devel] [imglib2] remove redundant KDTree implementation in imglib2-algorithms?

[Lee Kamentsky]

Thanks Tobias.

On 2/6/2012 12:25 PM, Tobias Pietzsch wrote:
> Hi guys,
>
> I was just about to use imglib2 KDTree and I noticed that there
> is still the legacy algorithm.kdtree package inherited from imglib1.
>
> There are now two implementations of a KDTree in imglib2.
>
> One is is the legacy imglib1 KDTree in net.imglib2.algorithm.kdtree.
>
> The other one is in the core.
> KDTree itself is in package net.imglib2.collection.
> The n-nearest-neighbor and radius search on this KDTree are in the
> net.imglib2.neighborsearch package.
>
> The new KDTree does not require it's T to be a Leaf< T >
> and the searches implement the appropriate 
> net.imglib2.neighborsearch.*NeighborSearch< T > interfaces.
>
> The new KDTree uses exact median-finding algorithm for constructing the
> tree instead of the approximate one in imglib1 kdtree.
> And it is faster.
>
> Because this is a confusing situation, _I would like to remove the
> "old" KDTree from the imglib-algorithms package_.
>
> (I would fix Lee's VolumetricSearch implementation which is the only
> thing using the imglib-algorithms KDTree currently.)
>
> Any objections?
>
> best regards,
> Tobias
>
> _______________________________________________
> ImageJ-devel mailing list
> ImageJ-devel at imagej.net
> http://imagej.net/mailman/listinfo/imagej-devel