cpfind - add option to limit number of control points

Cpfind adds with default setting 50 cp per image pair. If you want to change these numbers use the switches sieve2width, sieve2height and sieve2size as described in the manual http://wiki.panotools.org/Cpfind#Feature_matching

Hello,

I really do not understand what sieve2width, sieve2height and sieve2size do. The description is not very helpful to people who do not know (or care ;-) about the algoritm and its implementation. What combination of parameters do I have to set to get e.g. 7 cp-points per image pair?

And if there is some recipe for this, could it be wired to the "Points per overlap" setting in hugin's Images tab? [ I can move this sub-issue off to a second bug report ("With cpfind the "Points per overlap" setting in hugin's Images tab is ignored") if you would prefer that.

thanks, cu andreas

The sieve2* switches are used to get a better distribution of the cps. As far as I unterstood the image is divided in sieve2width * sieve2height parts/rectangles. In each of it cpfind keeps only sieve2size control points.

If you want to use the setting on the image tab add "--sieve2width 1 --sieve2height 1 --sieve1size %p" to the cpfind parameters. But then you will get a worser distribution of the cp.

I'm not sure if a complete support of all 3 parameters is really needed in the GUI. But for further investigations set to wishlist.

On 2010-12-05 tmodes wrote:
> The sieve2* switches are used to get a better distribution of the cps.
> As far as I unterstood the image is divided in sieve2width *
> sieve2height parts/rectangles. In each of it cpfind keeps only
> sieve2size control points.

Thank you for the explanation. That clears things up.

> If you want to use the setting on the image tab add "--sieve2width 1
> --sieve2height 1 --sieve1size %p" to the cpfind parameters. But then you
> will get a worser distribution of the cp.

Ok. I will set --sieve2width 3 --sieve2height 3 --sieve1size %p and
keep p at something like 2 or 3.

However I would still *like* to have my original wish: "Keep only the
best ten control points", somehow magically balancing quality and
distribution.

> I'm not sure if a complete support of all 3 parameters is really needed
> in the GUI.
[...]

I am quite happy with current single control for this setting, it
just should work.

thanks for your assistance.
cu andreas

Kay has a Python script to do what you want, Andreas.

http://groups.google.com/group/hugin-ptx/browse_thread/thread/ee15b52bbd82fc97

I agree that it would be a nice feature to have in the Hugin GUI to keep only the best N control points, somehow balancing quality and distribution.