Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 16:07 GMT
|
|
Hi!
I worked with panos for a long time, but now I got some new equipment and can't set it up to make (very) good panos. There are still stitching errors, but I gave my best (hopefully) to locate NPP.
Atm I do use: 5D II with shaved Tokina 10-17 @12mm and NN R1.
R1 is set to 2.4 at 0°, doing 6 shots (better results than 4 shots).
The Lens corrections parameters PTGui shows me are:
a: -0,22005468
b: 0,19017942
c: -0,087680767
On the NN Forum (which isn't very active) I found a thread (www.nodalninja.com/forum/showthread.php?3716-R1-s...) someone providing test images.
Loading these images in PTGui gives me a perfect pano, with loads of controls points from 0.1 to 2.0 and abc parameters at 0,02... 10 times betters!
I do have some good control points, too. But most of them are around 5 and up to 15+.
Do I have problems due still having the wrong NPP? Or might something else be the reason?
Greetings,
Florma
|
|
alert moderator
|
|
Doug Aurand
Posts: 3282
Location: Albuquerque, NM, United States
Registered: 2 Jan 2008
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 16:25 GMT
|
|
Florma
When you say there are stitching errors, are there visible misalignmnets?
Or are you assuming that a control points with larger numbers are an "error"?
I sometimes get a control point variation up to 8 or 9 with no visible misalignment. Other times I add extra control points to an area with a visiblbe misalignmnet, get an alignment to 1 and only get a modest improvement in what I can see.
For those instances, I just edit the misalignment using Pano2VRs Patch Tool or edit the Equirectangular Projection tiff directly in Photoshop
Can you post the image you're talking about so we can see what you're seeing?
Doug Aurand
Albuquerque, NM
|
|
alert moderator
|
|
bigwade
Posts: 821
Location: Utrecht, Netherlands
Registered: 19 Oct 2005
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 16:58 GMT
|
|
I use my R1 lensring setting about 2.55
But with six around it doesn't matter that much.
But eh..... show an image of the set(ting).
I have heard some people mounted it the other way........ 
|
|
alert moderator
|
|
Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 17:02 GMT
|
|
I hope you don't bother looking at my toilet, but I found this the most complex room in my apt due to the tiles
pano123.s3.amazonaws.com/Archiv.zip
|
|
alert moderator
|
|
Hans Nyberg
Posts: 2760
Location: Denmark
Registered: 28 Aug 2005
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 17:10 GMT
|
|
First of all 6 shots is worse than 4.
It's unfortuantelly a misunderstanding among people that they get better stitching with more shots using a fisheye.
Remember that the more shots you take the larger the difference in NPP along the blending area becomes.
The Tokina has an NPP which differs from 92 to 98 depending on the angle and by using 6 shots you include all angles.
Take 4 images. I can not give you the numbers for the R1 as I do not use this kind of panoheads. But from camera thread to rotation I use 96. That gives me a = -0.18 b = 0.08 c = 0 (I never optimize for c) FOV depends on your cropcircle it can be from 182-187 if you have done some small adjustments.
Remember to optimize for horizontal and vertical shift. In theory you should not need it if you adjust the crop circle exactly to the edge but that does usually not work in practice.
The Tokina can be difficult to optimize and get perfect result at the zenith just like the shaved Nikkor 10.5 and therefore I always take a Zenith when it needs to be perfect.
Thats also why I would never use one of these ring panoheads.
Hans
|
|
alert moderator
|
|
Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 17:12 GMT
|
|
Misaligments as images don't match, especially tiles.
As mentioned, I'm very impressed with the images shown in the thread. The do align perfectly.
I tried a 33° slant with 15mm, too. But stitching ended in a mess 
|
|
alert moderator
|
|
Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 17:17 GMT
|
|
Thanks for clarifying 6 vs 4.
Reason for 6 is that the bubble level of R1 matched better than with 4, which (for me) was an indicator 6 might be the better choice.
I'll try with 4 again later.
|
|
alert moderator
|
|
Hans Nyberg
Posts: 2760
Location: Denmark
Registered: 28 Aug 2005
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 17:41 GMT
updated: 24 Aug 2010 at 17:45 GMT
|
|
They stitch perfect. I delited quite a lot of controlpoints and ended with 2,5 and max 6.4.
As far as I can see there is no parallax but you will get some large differences in controlpoints due to using 6 images.
You will need to delete a lot of them to have controlpoints where they give exact match.
Here is the stitch www.panoramas.dk/test/Archivenyberg.zip
Oh yes I had to ad some vertical controlpoints.
Another problem with 6 images is that you get controlpoints between for example image 1 and 3 and that may often cause errors and even drag the whole panorama into a non level state.
Hans
|
|
alert moderator
|
|
Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 17:57 GMT
|
|
Thank you so much! With shift options and not use c I do get the results for myself.
Already tried 4 images with R1 at 2.55 but this does not work out.
I would like to use 4 as less work, but I'm happy with the results my 6 provide now.
Thanks a lot!!!
|
|
alert moderator
|
|
Hans Nyberg
Posts: 2760
Location: Denmark
Registered: 28 Aug 2005
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 18:16 GMT
|
|
Florma said: Thank you so much! With shift options and not use c I do get the results for myself.
Already tried 4 images with R1 at 2.55 but this does not work out.
I would like to use 4 as less work, but I'm happy with the results my 6 provide now.
Thanks a lot!!!
Why would it not work. You get enough overlap at 12.
You may get a little less controlpoints generated but if you make a template with the correct settings for the lens you do not need many.
Also if PTGui has all these settings it generates much more and better controlpoints.
This is what NN has for it.
Tokina 10-17mm f/3.5-4.5 DX for Canon and Nikon
4 shots around at 12mm (shaved lens on full frame sensor)
Tilt angle NPP value
+7.5° 1.55
+5° 1.90
0° 2.55
-2.5° 2.90
-7.5° 3.55
It looks like the setting changes depending on the tilt
Hans
|
|
alert moderator
|
|
Smooth
Posts: 3729
Location: Mount Panorama, Australia
Registered: 21 Jul 2004
|
|
Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
|
Florma
Posts: 9
Location: Germany
Registered: 24 Aug 2010
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 19:10 GMT
|
|
Well, I've to say thanks a lot for all your help.
|
|
alert moderator
|
|
bigwade
Posts: 821
Location: Utrecht, Netherlands
Registered: 19 Oct 2005
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 20:29 GMT
|
|
"First of all 6 shots is worse than 4."
Not, because you can use the best part of the lens.......
and with 4 you get a blurred star at nadir for sure.
With 6 the star is ±less.
BTW the Tokina is rather crap @12 on a FF at edges so you have allways problems with some kind of a blurred star in nadir.
It's designed for a crop camera and does a perfect job for those kind of camera's.
"The Tokina can be difficult to optimize and get perfect result at the zenith just like the shaved Nikkor 10.5 and therefore I always take a Zenith when it needs to be perfect.
Thats also why I would never use one of these ring panoheads."
You should try before you judge..
The R1 is a very, very nice and versatile head for travelling and more.
|
|
alert moderator
|
|
Hans Nyberg
Posts: 2760
Location: Denmark
Registered: 28 Aug 2005
|
Re: Still stitching errors, but located NPP as good as possible
Posted: 24 Aug 2010 at 21:18 GMT
|
|
If YOU get a blurred star at nadir you must also get it at zenith,
I don*t
At nadir I have the rotator so I can not see anything.
Hans
|
|
alert moderator
|
|
back to threads list |
this thread is closed
Prev
