Image-stitching and virtual tour solutions My account Updates
It is currently Fri Oct 31, 2014 5:21 am

All times are UTC + 1 hour




Post new topic Reply to topic  [ 5 posts ] 
Author Message
PostPosted: Sun May 10, 2009 4:26 pm 
Offline
Member

Joined: Wed Nov 14, 2007 2:12 pm
Posts: 14069
Location: Isleham, Cambridgeshire, UK.
I've spent more time looking at the efficiency/performance of the Papywizard Import filter and find some strange behaviour with CP detection.

It sometimes seems to fail to detect CPs on image pairs where you would expect it to find good CPs.

The test image set shot using Merlin/Papywizard, Nikon D40 with kit zoom 18-mm set at 35mm (52.5mm equiv.) at lowest res (about 1MP) and basic quality JPEG. A mosiac of 9 rows and 10 columns - 90 images in total.

Image set available here:
http://www.three60views.org.uk/image_sets/merlin-mosaic-front-90.zip (18MB ZIP)

Screenshot 1 - shows image placement.

Screenshot 2 - shows portion of rendered (spline36/multiband) pano displaying obvious stitch errors on car.

Screenshots 3 and 4 - shows Links made during import/detection - missing links between image adjacent images in area of the car and roof of neighbouring house.

Screenshots 5-7 - show image pairs where program failed to detect CPs, or failed to place CPs where you would expect, but there appears to be no obvious reason why not.

Screenshots 8-10 - shows that manual CP detection worked in every 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.


Last edited by mediavets on Sun May 10, 2009 4:39 pm, edited 1 time in total.

Top
 Profile  
 
 Post subject:
PostPosted: Tue May 12, 2009 1:32 pm 
Offline
Administrator
User avatar

Joined: Mon Nov 14, 2005 4:56 pm
Posts: 5901
Location: Francin, France
Ok, this one owns a full detailed explanation. How does the detection / optimization work when using papywizard :

Detection :
- Each camera read from the xml file.
- From these coordinaets, we can precalculate a potential list of image pair which are possible ( 1<->2, 1<->16, 1<->17 is possible but no other link for picture 1, etc )
- All CP are extracted and compared by using this list of possible pair
- The panorama is displayed with those CPs
Optimization : all algorithms in this part are now applied.
- first, as we are using grid optimization, the initial position of all images is given by the xml file.
- having a not so bad initial location of every camera, we can calculate RMS for each link.
- by default, the preset tells to do a cleaning of pair according to that current quality (RMS) ( rms calculated on the CP but image have been located by xml file only, not by the cp ).
The current setting tells to prune wrong link over 20 RMS. In practice, I've notice that this value is too low and also cleans real link. It's the case here.
- then, we do the real optimization and you'll get the good panorama.

Knowing that, you can try this experience :
- use the "skip optimization" flag in the papywizard to see the rms quality before optimization ( calculated just from the camera location recorded in xml ).
- uncheck "auto detection" and in group setting, remove the clean before final optimization. No link will be cleaned.


Top
 Profile  
 
 Post subject:
PostPosted: Tue May 12, 2009 4:29 pm 
Offline
Member

Joined: Tue Aug 29, 2006 8:24 pm
Posts: 91
Location: nyc
I have found that after geometry analysis optimizing with the default "gigapixel" setting will delete every CP.

The import tool does not seems to update the grid from a new "central position" when using images at 1000mm focal length. After optimizing the grid positioned images (no CP) are not even close to the other adjacent images.








Top
 Profile  
 
 Post subject:
PostPosted: Tue May 12, 2009 4:47 pm 
Offline
Member

Joined: Wed Nov 14, 2007 2:12 pm
Posts: 14069
Location: Isleham, Cambridgeshire, UK.
AlexandreJ wrote:
Knowing that, you can try this experience :
- use the "skip optimization" flag in the papywizard to see the rms quality before optimization ( calculated just from the camera location recorded in xml ).
- uncheck "auto detection" and in group setting, remove the clean before final optimization. No link will be cleaned.

Alexandre

Thank you for your explanation of what the Papywizard Import filter does and how it works.

Your suggested alternative approach/'experience' worked VERY well producing a perfect result - perhaps that CP cleaning parameter in the Papywizard Import filter needs some adjustment as you say.

Compare the screenshots below with similar screenshots in this post:
http://www.autopano.net/forum/p44941-2009-05-10-16-26-21#p44941







_________________
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.


Last edited by mediavets on Tue May 12, 2009 4:55 pm, edited 1 time in total.

Top
 Profile  
 
 Post subject:
PostPosted: Tue May 12, 2009 5:40 pm 
Offline
Administrator
User avatar

Joined: Mon Nov 14, 2005 4:56 pm
Posts: 5901
Location: Francin, France
BTW. I added a flag in option tab of the papywizard to activate this cleaning or not. By default, it will be off, because this system seems to create more issue than solving real problem.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 5 posts ] 

All times are UTC + 1 hour


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group