Different FOV's -> Weird estimations

[rew]

I have a 12 degree FOV image and a 360 degreen FOV image. Hugin doesn't estimate the positions of additional control points correctly.

I've found that if I select the second control point from the big image (the 360 degree one), I will get: estimated control point outside of image. This means that hugin won't pan the images around with me losing sight of the point I wanted to add.... (the big image needs to be viewed at 100% to be able to see details just large enough to see on the small images).

[Bruno Postle (brunopostle)]

This is a known bug, related to bug #1979803 and bug #1979796. Basically all the auto control point tools assume images of the same lens type.

[zarl (zarl)]

It looks like hugin is used to synch a set of tele shots with a low resolution background. Following Bruno's remark I would suggest to first take the 360 deg. shot alone and export a set of 12 deg. rectilinear frames that show the vicinity of the high resolution shot. Take those exported frames to find your CPs. Does that work?

[rew (r-e-wolff)]

It works, but it's cumbersome. In this case I'm not "fishing" for a workaround, but hoping that hugin could be improved in the future.

[rew (r-e-wolff)]

In general the hugin control point editor can be improved.

- Allow rotating of the images through the gui (not only through running an optimizer step, and then having the gui detect which side it thinks is up.

- Allow non-90 degree rotations to line things up. (I've been known to pick the wrong control point when SIFT didn't work either).

- Allow different scales. Just a linear scale will do. (in the case above the 12 degree shots are 1/30th of the 360 degree pano.) (The example above is a bit extreme, but I really want to be able to make those panos eventually).

[rew (r-e-wolff)]

 Bug #679871

[Yuv (yuv)]

Located the two reports mentioned by Bruno in comment #2 and marked them as duplicate of this one to make the link. Also marked the bug identified by Rogier in # as duplicate, to collect in a single place (here) all that is known to be caused by the limitation of auto CP generation.

Does this affect also the new cpfind?

[rew (r-e-wolff)]

Yuv, I was not talking about auto-cp-finders. It's the interactive tool I started all this about.

[tmodes (tmodes)]

Fixed in default branch. Auto-estimate takes now the projection and fov into account.