[ImageJ-devel] [fiji-devel] Fwd: Local Neighborhood stuff

[Barry DeZonia]

Hi guys,

I've been too busy to dig very deeply into this. I just wanted to let you
know I dealt with some similar issues when writing the
net.imglib2.ops.pointset package. Giving it a look might be informative.

One thing that jumps out at me is the reliance on a span for an abstract
neighborhood. I think there should be bounds but that neighborhoods might
be sparse and possibly nonsymmetric. Imagine a 2d neighborhood like this:

X  X  X  X
    X
X  X

And let any point could be the "center" (or anchor) depending upon the API
user's wishes. This is the idea behind GeneralPointSet. If span means bound
then no problem but if span means "all the points within this n-d space"
then the above is not supportable.

Anyhow I hope this adds to the discussion. If there are any questions in
regard to the PointSet package let me know.

On Fri, Jul 27, 2012 at 10:35 AM, Jean-Yves Tinevez <
jeanyves.tinevez at gmail.com> wrote:

>
>
> Hi all
>
> I pushed some changes to the test branch: I removed the realpositionable
> stuff, and tried to draft the mother interface from our discussions.
>
> But I got stuck at the copyOn method. If I try to make it as generic as I
> can (copy on a different RandomAccessible type and on a different type),
> concrete implementations that are more specific (like the buffered
> rectangle) fail on bad type bounds. I could use some help there.
>
> Best
> jy
>
>
>
>  --
> Please avoid top-posting, and please make sure to reply-to-all!
>
> Mailing list web interface: http://groups.google.com/group/fiji-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://imagej.net/pipermail/imagej-devel/attachments/20120731/072631ac/attachment.html>