Aligning horizontal and vertical lines  

Share your tips and tricks here or get help with any Autopano Pro / Giga problem!
No bug reports (of any kind) in this forum!
no avatar
knmiller@me.com
Member
 
Topic author
Posts: 15
Likes: 0 post
Liked in: 0 post
Joined: Mon Jan 03, 2011 5:37 pm
Info

Aligning horizontal and vertical lines

by knmiller@me.com » Sun Feb 13, 2011 8:09 pm

Hi I have shot some interiors which have strong horizontal and vertical lines. In a few places the lines from one frame to another have not been perfectly aligned. I have used the correct settings for my camera, lens and head combination (I have tested these settings myself for parallax error), so can only assume I need to be doing something extra in Autopano pro to get all the lines to align perfectly.

What would you recommend I try to correct this problem, I have attached a picture to illustrate the issue.

Thanks

Kevin







no avatar
mediavets
Moderator
 
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

by mediavets » Sun Feb 13, 2011 9:41 pm

What camera/lens and pano head are you using?

What pattern of shots have you used?

May we see a screenshot of the complete pano scene to provide context?
Andrew Stephens
Many different Nodal Ninja and Agnos pano heads. Merlin/Panogear mount with Papywizard on Nokia Internet tablets.
Nikon D5100 and D40, Sigma 8mm f3.5 FE, Nikon 10.5mm FE, 35mm, 50mm, 18-55mm, 70-210mm. Promote control.

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Sun Feb 13, 2011 10:09 pm

knmiller@me.com wrote:What would you recommend I try to correct this problem, I have attached a picture to illustrate the issue.

Hi Kevin!

I guess you should review your camera/lens/head alignment carefully - those seem to me to be typical errors caused by misalignment. As Andrew said already it would be wise to give more informations.


best, Klaus
Last edited by klausesser on Sun Feb 13, 2011 10:13 pm, edited 1 time in total.
Simplicity is the keynote of all true elegance. Coco Chanel

User avatar
foto-chemnitz
Member
 
Posts: 22
Likes: 0 post
Liked in: 0 post
Joined: Fri Apr 24, 2009 6:33 pm
Location: Chemnitz - Germany
Info

by foto-chemnitz » Mon Feb 14, 2011 11:57 am

Hello Guys,
I often have the same prolems as Kevin.
Dont belive (in every case) in a bad lens/camera/head combination.
Now with APG 2.5 (antighost) I have worse results than with APG 2.0.9 (smartblend).

I have uploadet some test-files: www.foto-chemnitz.de/downloads/autopano_test.zip
Can somebody try to stitch this with APG 2.5 and tell me/us the used settings. Iam very thankful for such test...
I tryed a lot with manual control points, but without much success :-(

Jan


Last edited by foto-chemnitz on Mon Feb 14, 2011 12:01 pm, edited 1 time in total.

no avatar
Destiny
Moderator
 
Posts: 7886
Likes: 6 posts
Liked in: 228 posts
Joined: Mon Apr 26, 2010 11:55 pm
Info

by Destiny » Mon Feb 14, 2011 1:07 pm

if you say your setting are all fine then I think your tripod moved... I tried to stitch it even trying PTGui Pro.. same thing.. ether that or your npp is not set right.. it looks to me that your camera is offset to the right... but the global RMS are excellent.. I would like to see that on my stitching but I struggle to get 4 on the other hand my VR's look great... I do not think that manual control points will help... it would pull other parts of the image out.. I then put it into Photoshop Automate Photomerge, you can see the image being pulled in from the left to the right.. You camera/lens/tripod etc therefore are not set right...
http://members.westnet.com.au/dps/Panorama1Destiny.html

Hope this all helps...

D

no avatar
GURL
Member
 
Posts: 2943
Likes: 0 post
Liked in: 0 post
Joined: Tue Dec 06, 2005 1:57 pm
Location: Grenoble
Info

by GURL » Mon Feb 14, 2011 1:52 pm

Destiny wrote:if you say your setting are all fine then I think your tripod moved... I tried to stitch it even trying PTGui Pro.. same thing.. ether that or your npp is not set right.. it looks to me that your camera is offset to the right... but the global RMS are excellent..
D

My own results are very very very unusual.

I cant find any parallax change when looking at the source images.

Using 2.5 Final I got a good Global RMS with some very bad CP but:
- source images are matching only next to the center of the overlap region
- focal length value displayed in Layers sub-window is stupid (37 mm instead of the 17 mm value Autopano found in EXIF and EXIFtool confirmed.)

My bet: not a Multiband versus Smartblend issue. This could be related with ultra-wide-angle lens having more problems than they usually have. A bug or glitch in this particular 2.5 Autopano version?
Last edited by GURL on Mon Feb 14, 2011 1:56 pm, edited 1 time in total.
Georges

no avatar
GURL
Member
 
Posts: 2943
Likes: 0 post
Liked in: 0 post
Joined: Tue Dec 06, 2005 1:57 pm
Location: Grenoble
Info

by GURL » Mon Feb 14, 2011 3:11 pm

By deleting some auto CP and adding a bunch of manual ones so that they were well spead along the overlap, I was able to make Autopano using a more appropriate focal length (18.3 mm)

The stich is decent but there is an error visible at the top. Contrary to some opinions my guess is that the new Multiband "Remove ghosts" option is helping a lot at masking optimization errors.

When CP are located or mostly located next to the image center Autopano optimizer can't find the proper lens distortion and focal length ? If this is the problem, having good lens profiles for UWA like this one could help...

BTW, too many of us tend to bielieve in global RMS value rather than watching the preview. Global RMS is very helpfull for beginners but watching a zoomed preview is much more accurate.




Last edited by GURL on Mon Feb 14, 2011 3:29 pm, edited 1 time in total.
Georges

no avatar
GURL
Member
 
Posts: 2943
Likes: 0 post
Liked in: 0 post
Joined: Tue Dec 06, 2005 1:57 pm
Location: Grenoble
Info

by GURL » Mon Feb 14, 2011 3:54 pm

Here is another attempt using 2.5:

- lens correction unselected
- no individual CP added
- CP clouds added near the top and bottom

Though the error next to the ceiling is still there, unselecting Lens Correction option should help (for UWA lens only.)


Last edited by GURL on Mon Feb 14, 2011 3:57 pm, edited 1 time in total.
Georges

no avatar
mediavets
Moderator
 
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

by mediavets » Mon Feb 14, 2011 7:11 pm

GURL wrote:My bet: not a Multiband versus Smartblend issue. This could be related with ultra-wide-angle lens having more problems than they usually have. A bug or glitch in this particular 2.5 Autopano version?

This is a fairly 'unforgiving' pano scene, and APP/APG has always had more problems with very wide angle rectilinear lesnses - compared to other types of lens - I think.

It appears that this is may still be the case?
Andrew Stephens
Many different Nodal Ninja and Agnos pano heads. Merlin/Panogear mount with Papywizard on Nokia Internet tablets.
Nikon D5100 and D40, Sigma 8mm f3.5 FE, Nikon 10.5mm FE, 35mm, 50mm, 18-55mm, 70-210mm. Promote control.

no avatar
GURL
Member
 
Posts: 2943
Likes: 0 post
Liked in: 0 post
Joined: Tue Dec 06, 2005 1:57 pm
Location: Grenoble
Info

by GURL » Mon Feb 14, 2011 8:21 pm

Destiny wrote:I tried to stitch it even trying PTGui Pro.. same thing..

Comparisons are difficult, many different ways can be followed during optimization, etc.

mediavets wrote:APP/APG has always had more problems with very wide angle rectilinear lenses - compared to other types of lens - I think.

A question I sometimes asked and nobody ever answered is whether ultra-wide-angle rectilinear lenses having the same focal length than a given fisheye would give better, equivalent or worst result when used for stitched panos. They are both used to build a virtual sphere. Rectilinear lenses are using (wasting ?) a large amount of pixels to record the areas of the sphere corresponding to their four corners. As a result I believe that, for an identical image number, a fisheye is preferable but, because rectilinear UWA require more source images than a fisheye, this argument is not very conclusive...

Another point is that UWA often show very complex distortions: when, like in this example, CP are located next to the image centers, a good distortion correction is not possible. Being able to forbid Autopano attempts to adjust the focal length could help, perhaps ?
Georges

no avatar
Destiny
Moderator
 
Posts: 7886
Likes: 6 posts
Liked in: 228 posts
Joined: Mon Apr 26, 2010 11:55 pm
Info

by Destiny » Mon Feb 14, 2011 10:10 pm

Such a good point... I have seen mostly 'good', never 'excellent' global RMS on my image stitching results, however the preview itself in my eyes is 98 percent perfect... Infect, hard to find faults unless looking... Nevertheless, I am always striving to improve... The forum on the whole has be very positive and supportive in solving problems and issues we all have... 'Unless you drive little red cars' ;)

D

BTW, too many of us tend to bielieve in global RMS value rather than watching the preview. Global RMS is very helpfull for beginners but watching a zoomed preview is much more accurate.

no avatar
hankkarl
Member
 
Posts: 1284
Likes: 0 post
Liked in: 0 post
Joined: Tue Feb 21, 2006 5:32 pm
Location: Connecticut, USA
Info

by hankkarl » Tue Feb 15, 2011 2:34 am

GURL wrote:
Destiny wrote:I tried to stitch it even trying PTGui Pro.. same thing..

Comparisons are difficult, many different ways can be followed during optimization, etc.

mediavets wrote:APP/APG has always had more problems with very wide angle rectilinear lenses - compared to other types of lens - I think.

A question I sometimes asked and nobody ever answered is whether ultra-wide-angle rectilinear lenses having the same focal length than a given fisheye would give better, equivalent or worst result when used for stitched panos. They are both used to build a virtual sphere. Rectilinear lenses are using (wasting ?) a large amount of pixels to record the areas of the sphere corresponding to their four corners. As a result I believe that, for an identical image number, a fisheye is preferable but, because rectilinear UWA require more source images than a fisheye, this argument is not very conclusive...

Another point is that UWA often show very complex distortions: when, like in this example, CP are located next to the image centers, a good distortion correction is not possible. Being able to forbid Autopano attempts to adjust the focal length could help, perhaps ?

Having a 12-24mm Sigma on a 5D and no fisheye, I can't say for sure. But I think the rectilinear lens distorts the image, and then APP has to undo the distortion.

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Tue Feb 15, 2011 10:41 am

I tested this case. It's a bit strange but can be explained.
First, a standard detection with default settings. Not really nice because of the lack of CP overall on the overlapping zone.
I raised the number of CPs, it didn't changed anything => a bug is lying here because it seems that the cp slider isn't used ...
So, I manually added more cp on the corners of the overlapping zone. It does help but not that much.
To get a good result, I needed to activate multiple viewpoint. With that setup, the stitch is quite good ( except on the top and really bottom ).
My guess is that both images were not really shot at the nodal point.







User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Tue Feb 15, 2011 1:40 pm

GURL wrote:When CP are located or mostly located next to the image center Autopano optimizer can't find the proper lens distortion and focal length ? If this is the problem, having good lens profiles for UWA like this one could help...

UWA lenses having non-linear distortions are hard to compensate for all angles - as you mentioned already. If this issue meets a lign-up which is not 150% perfect in such a small environment . . . .
Again: i suggest to 1) check the NPP settings VERY carefully and 2) check the lenses distortion pattern. You can download a dedicated grid pattern and photograph it with the lens you use for panos. Then you use a lens-correction application (Adobe´s isn´t bad and it´s free) to compensate distortions. The resulting file you should use for every shot (!) you take with this lens before (!) stitching the pictures.

A more comfortable way would be to get a lens with lesser distortions and check the line-up VERY carefully, when shooting in small rooms at close distances . . . because that´s really challenging.

But maybe the lens is ok and it´s related to trhe settings - it´s difficult to set it prtecisely for all used angles for use in narrow spaces and difficult wall-patterns.

On the other hand it´s very easy to retouch such errors.

best, Klaus
Simplicity is the keynote of all true elegance. Coco Chanel

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Tue Feb 15, 2011 2:08 pm

GURL wrote:Another point is that UWA often show very complex distortions: when, like in this example, CP are located next to the image centers, a good distortion correction is not possible. Being able to forbid Autopano attempts to adjust the focal length could help, perhaps ?

Yes - that´s a vital point and a problem using UWA´s of not first-class quality. Which are rare and very expensive. A good way to compensate such complex distortions is to downlad a correction-pattern, photograph it using a precise ligning-up and let a lenscorrection application calculate a compensation - that should be done at EVERY aperture which is usually used for the shootings (i myself use always the same aperture - the one which shows the best optical performance, so that it´s not a too heavy task).
Having photographed the correction-pattern a compensation in a lens correction application produces a corection-file which can compensate even very complex distortions.

Apps like DXO, Bibble, Lightroom or so can compensate to some extent (IF it is ALL in the EXIFs) - but how far they go checking all lenses or what they get as informations from the vendors nobody knows . . so i prefer to check such things myself in the case an issue occurs :cool:

best, Klaus
Last edited by klausesser on Tue Feb 15, 2011 2:16 pm, edited 1 time in total.
Simplicity is the keynote of all true elegance. Coco Chanel

no avatar
GURL
Member
 
Posts: 2943
Likes: 0 post
Liked in: 0 post
Joined: Tue Dec 06, 2005 1:57 pm
Location: Grenoble
Info

by GURL » Tue Feb 15, 2011 6:52 pm

In a new attempt: I removed many auto-CP and added manual ones from the top to the bottom of the overlapping area. Using 100 CPS (BTW manual CP tool works great !) I reached a global RMS of 1.9 where the worst CP RMS was 3.

Practical conclusion is clear: Multiple View Points and Anti-ghosts are to be used...
... but the problem remains very cloudy as using no Anti-ghosts shows that the source images don't match well even with good CPS and good global RMS.

As far as I can see the NPP is well located and the tripod was not moved. Whether the zoom setting was changed or not I don't know.

This looks like an excellent example about what Anti-ghost can do (removing ghosts :) ) and can't do (avoiding broken lines :( )


Last edited by GURL on Tue Feb 15, 2011 6:55 pm, edited 1 time in total.
Georges

User avatar
foto-chemnitz
Member
 
Posts: 22
Likes: 0 post
Liked in: 0 post
Joined: Fri Apr 24, 2009 6:33 pm
Location: Chemnitz - Germany
Info

by foto-chemnitz » Wed Feb 16, 2011 12:27 am

Hello,
thank you very much for all your tests!!!! Very informative!

I forgot to say that I corrected the lens distortions in Photoshop Lightroom before stitching.

Now I used the original uncorrected pictures with much better results.
Does somebody know the reason for this?
Maybe a double correction (first in Lightroom, second in Autopano) is not good???

Greetings from Germany

no avatar
GURL
Member
 
Posts: 2943
Likes: 0 post
Liked in: 0 post
Joined: Tue Dec 06, 2005 1:57 pm
Location: Grenoble
Info

by GURL » Wed Feb 16, 2011 2:01 am

Lightroom used some distortions of its own to counterbalance the lens distortions but :o Autopano don't know how to undistort such distorted distortions!
Georges

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Wed Feb 16, 2011 9:05 am

Yes GURL, that could be an explanation. If you have original image, we could test that.

User avatar
foto-chemnitz
Member
 
Posts: 22
Likes: 0 post
Liked in: 0 post
Joined: Fri Apr 24, 2009 6:33 pm
Location: Chemnitz - Germany
Info

by foto-chemnitz » Wed Feb 16, 2011 11:46 am

The thing with the distorted distortions sounds comprehensible.

I uploaded the original files here: www.foto-chemnitz.de/downloads/panotest_undistort.zip

Greetings - Jan -
Last edited by foto-chemnitz on Thu Feb 17, 2011 3:24 pm, edited 1 time in total.

no avatar
rmfrance
New member
 
Posts: 3
Likes: 0 post
Liked in: 0 post
Joined: Mon Jan 10, 2011 4:25 pm
Info

by rmfrance » Tue Feb 22, 2011 3:07 pm

A very interesting discussion! Using Autopano Pro V2.09 on some images with parallax-induced alignment problems, I tried unchecking the Lens Distortion option and can confirm that (in my case, at least) I then achieved much better stitching than previously.
Lens-wise, I use a Tokina 11-16 ATX-Pro S f2.8, which I found to be among the best performers available in terms of distortion.


Who is online

Users browsing this forum: No registered users and 2 guests