[MINC-users] mincdefrag arguments

[Alex Zijdenbos]

OK - I may be able to get mincdefrag functionality incorporated into
mincmorph (may take a little while).

For the short term, I'll fix the mincdefrag help text and command-line
checking this week; will keep it backward compatible as I suggested. I
prefer (6, 18, 26), since we are talking about #neighbors.

Silently doing something else is not just undesirable, it's wrong :)

-- A


On Mon, Jun 23, 2014 at 11:37 PM, Claude LEPAGE <claude at bic.mni.mcgill.ca>
wrote:

>
> hi,
>
> Somehow, mincdefrag survived over the years after I implemented it quickly.
>
> I have no objection if someone wants to include it in mincmorph. That's
> fine,
> as long as that someone is not me.
>
> I agree that the specification of the arguments is not consistent with
> other
> minctools. Maybe 6 should be 7 instead. But it's all the same. The center
> voxel is always included. The "silent" mode is perhaps not desirable. It
> would
> not take much effort to make it "idiot-proof".
>
> Claude
>
> PS: It's nice to see that mincdefrag is being used out there. :-)
>
>
> > IIRC, there was at one stage an effort to include the logic of
> > mincdefrag into mincmorph. (it would be just another kernel
> > operation). This would be one way to resolve differences and allow
> > mincdefrag to be more flexible.
> >
> > Claude, do you remember what happened here?
> >
> >
> > a
> >
> > On 24 June 2014 09:14, Alex Zijdenbos <zijdenbos at gmail.com> wrote:
> > > This one is mostly for Claude, but just in case it has affected
> anybody I
> > > figured I'd raise it here.
> > >
> > > The mincdefrag help text states that the allowable values for this
> "number
> > > of neighbors" value are (6, 19, 27), which is inconsistent (19 and 27
> > > include the center voxel, 6 does not).
> > >
> > > Where this matters is that I just realized that mincdefrag actually
> doesn't
> > > check these values, and *silently (ack) falls back to '6' for anything
> that
> > > is not 19 or 27*.
> > >
> > > I for one have used (by force of habit or logic or consistency with
> > > mincmorph) '26' in quite a number of mincdefrag calls, and I am now
> > > realizing that those were all silently using a 6-neighbor kernel :(
> > >
> > > I suggest we change the help text to read (1, 18, 26); and, for
> backward
> > > compatibility, change the bit where it checks this value to accept
> both 18
> > > and 19 for stencil=2; 26 and 27 for stencil=3, and throws an error for
> > > anything that is not (6, 18, 19, 26, 27);
> > >
> > > -- A
> > > _______________________________________________
> > > MINC-users at bic.mni.mcgill.ca
> > > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
> > _______________________________________________
> > MINC-users at bic.mni.mcgill.ca
> > http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
> >
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>
>