 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
I try to promote again this thread because it is really important to improve detectivity and quality of CP in the software.
We are going soon to work on that part and the more case we have, the best the quality at the end will be.
So, let me explain again what we need :
- image pairs
- no relation found even with high detection quality in control point editor ( or really hard to get something ).
Those images should still overlaps, but can have few details, or one image can be a little blurred, etc.
We need the images in full size without any rescaling, with full exif.
Thanks.
Offline
what about larger image sets, where everything is scrambled up ?
the iron structure of the brigde is so badly detected:
http://www.autopano.net/forum/p35116-20 … -28#p35116
Offline
That's out of topic for this case. We just want image pairs that are hard to stitch together.
I'm treating currently your case in the gigapan importer which is not related to that topic demand.
Now : the gigapan importer would benefit from your case too, if you can provide images to us using the ftp.
Offline
Hi , if you like I can shoot a series from our place where we live. This is at higher altitude and very often we are surrounded by low clouds, that means clouds between trees in the surrounding valley in a great panorama. I had quite often some trouble to stitch it when the clouds where all sort of the same density. Regards: Jaap
Offline
Sure don't hesitate to share any kind of such case.
Offline
I am having repeated issues stitching images that were created with DoF stacking software like Helicon Focus, this probably relates to the transformations applied to the images to scale them to the same actual focal lenght during the process of DoF stacking, are you interested in such specific cases too?
Cheers,
Bernard
Offline
bernardlanguillier wrote:
I am having repeated issues stitching images that were created with DoF stacking software like Helicon Focus, this probably relates to the transformations applied to the images to scale them to the same actual focal lenght during the process of DoF stacking, are you interested in such specific cases too?
Cheers,
Bernard
Hi Bernard!
I do this very often - never had a problem to stitch such pictures.
Alex: "We need the images in full size without any rescaling, with full exif."
What if there are no exifs at all? How much depends APG on them?
best, Klaus
Offline
If there's not a good detection due to bad overlap of pairs, I think the 3 following SHOULDs should done:
1.autopano should explore this cases of individual images
2.autopano user should persists that autopano stitch all of the individual images of one panorama project(even with bad overlap of pairs)
3.autopano should virtually change our camera location , and so it will move our pairs to the another view for the best overlap supplied(It will be an advanced property)
I'm not a software engineer and so I don't know these plans will be applicable.I only told my thoughts.
Mohammad
Last edited by mohammad233b (2010-01-16 22:49:54)
Offline
How about image-pairs where APP finds CP's in featureless sky, and misses what would seem to be obvious matching CP's in non-sky?
I had a thread related to that recently (month or so ago). I think I uploaded .jpg's.
The suspicion was sensor noise might be involved, but I have my doubts that was the cause.
Offline
Seems like APP has problems with a single row when the camera on the panohead is tilted quite a bit back. An example would be the top of a nearby tall building. With a moderate to ultra wide angle lens, there is a lot of key-stoning that APP has to figure out.
Would that be an example you would be interested in?
Offline
yes !
Offline
I've got a set of images that behave oddly. It's no an image-pair, but might be a good test case.
* First attempt. Had trouble with light pole sticking up into sky, and overexposed image. Also found CP's in featureless sky. f8 with reasonably clean sensor, so probably not sensor dust. Tried to supply "hint box", which didn't help. At least it didn't crash, which often happens with the "hint box".
* Redid images in ACR to darken all images making up pano, and it did much better. Light pole ok.
* Out of curiouslity, changed project setting to "High Detection Quality" and ~70 CP's per image pair. Poor stitch ... worse than first, but reported 1.89 RMS.
I've got the 2.0.7 trial, so I can't save projects.
HTH
Offline
I have a pano that no pano program I tested so far was able to stitch without massive errors.
A room with a white ceiling with white stucco on it, a dark brown parquet floor
and white walls with large areas with very little contrast.
Those programs failed: Panorama Factory 5.3, Microsoft ICE, Autopano Pro 2.06.
45 images (3 rows x 15 pictures)
Is this interesting?
-Frank
Offline
Yes, share the panoramas through our ftp : http://www.autopano.net/forum/t766-ftp-server
Offline
I will upload to you some.
Offline
I notice controlpoints are being detected very well now. Only places with trees and clouds moving in wind and clouds changing due to thermals cause me grief now.
This kind of change is really not in the scope of APx currently?
When features physically change there is no way to stitch.
Maby what is needed is a manual mode where approximate overlap can be specified and the sw then tries using global lense correct value and find minimum rms placement by trial and error?
I could uppload some of my changing clouds if needed but I am sure they are plentifull.
Offline
