 |
|
|
|
|
|
 |
| ||||||||
 |
 
|
|
|
|
|
|
|
|
|
| ||||||||
|
|
| 
| User list | Rules | You are not logged in.
Pages: 1 2
Hey Guys,
i spend some few days trying to perfect my NPP (no Parallax Point) and tried different settings with my gear (Canon 7D + Sigma 10-20 @10mm + NN4) the Lower Rail settings is 54mm and Upper Rail 100mm.
when doing a 360*180° (with 16 imgs; 4*-60°, 8*0°, 4*60°) i have some shifting on the ceiling.
I send the images to Nodal ninja forum, and a guy with PTGui stitched my panorama without problem and no shifting.
Is there some settings i need to adjust on APP ? or my NPP is perfectible ?
here are the images :
http://ge.tt/3MZYxoL
Gepetto2 files are from PTGui and the 54-100 are from APP.
I'm using APP 2.6.3, for the blending i'm using anti-ghost, bicubic
Thanks for your help.
Offline
Hi.. I had the same issues way back but it took me a long time to work out what was happening... Essentially, apg and PTGui stitch images differently so as a consequence you need to have two NNP settings... Some might argue that this is not the case but it is... Its not a small amount either but its not the centring of the lens that is the issue, since that will remain the same no matter what lens you use with the same camera.. Its the protrusion of the lens setting which needs to change according to the software you use... My VR Drive is extremely accurate, so quick to test this process. I found that there is about a 5mm difference between software. It frustrated me that I could achieve perfect stitching with PTGui but apg had issues.. I found myself changing the protrusion of the camera/lens to accommodate the different software. You can see a very visual difference after using my VR Drive when I use my PTGui settings and then stitch using apg. For some reason, it appears apg disregards an entire image altogether so makes for a terrible footprnt image.
You can see my footprint using the PTGui setting and what apg does from #14 at this thread...
http://www.kolor.com/forum/viewtopic.ph … 67&p=1
Destiny...
Offline
Hi Ed.. Well I would have thought that if i set the NNP to ensure good stitching results in apg that if ... PTGui could cope better that the stitching would be ok using the apg settings but its not to be.. I often get Good but never Excellent.. If I set my NNP for PTGui using my VR Drive, I get Excellent at all times and its perfect stitching.. Now, the nadir footprint on the other thread shows the two footprints to be different but so is the stitching.. If I set the VR Drive to achieve Excellent from PTGui, not only the nadir footprint from apg is very different but the stitching is out in MANY areas, you can see a little bit on this image.. I realised there was an issue only after I saw the nadir footprint from apg after I bought my VR Drive.... I have sent the images to Alex J at his request to find a solution but so far no feedback... Perhaps if I can find a solution to fixing the nadir footprint, using the same NNP settings I use for PTGui, then the stitching too might improve..
Destiny...
Offline
Destiny wrote:
Some might argue that this is not the case but it is...
Yes - indeed: this is not the case.
What is different is the kind of lens models.
There´s only ONE NPP - completely indepemdent from the stitcher-application.
Klaus
Offline
No.. not the case at all.. But I also realise you are set with your opinion so there is no chance of changing it.. But I know you are wrong... I shouldn't be but it is a fact... PTGui and apg interpret the NNP differently.. Same camera, same lens... with different results....
Destiny..
Last edited by Destiny (2012-08-11 12:18:43)
Offline
Jean-Phi38 wrote:
Hey Guys,
i spend some few days trying to perfect my NPP (no Parallax Point) and tried different settings with my gear (Canon 7D + Sigma 10-20 @10mm + NN4) the Lower Rail settings is 54mm and Upper Rail 100mm.
when doing a 360*180° (with 16 imgs; 4*-60°, 8*0°, 4*60°) i have some shifting on the ceiling.
I send the images to Nodal ninja forum, and a guy with PTGui stitched my panorama without problem and no shifting.
Is there some settings i need to adjust on APP ? or my NPP is perfectible ?
here are the images :
http://ge.tt/3MZYxoL
Gepetto2 files are from PTGui and the 54-100 are from APP.
I'm using APP 2.6.3, for the blending i'm using anti-ghost, bicubic
Thanks for your help.
The point is: PTGui prefers to seek points in geometrical structures. APG seeks point everywhere. So you might have CPs in the sky where you don´t have ANY visible (!) structure (but of course there must be something to identify - we just can´t see it 
).
That definitely makes APG a more universal stitcher - but it can have it´s drawbacks also.
When PTGui finds points on structures it´s very precise and renders fine and sharp images - sometimes sharper and more even colored than APG.
You´re on the "safe side" using both of them. I sometimes encountered - no, not of the 3rd kind, Destiny - issues in APG which PTGui solved easily.
That was when i photographed a building with very much windows: moving clouds mirrored in them and with taking 80 shots it´s ovious that they differed from
row to row very much.
APG didn´t get it that time - and i wasn´t experienced at all and at that time there was no manual CP setting in APP/G (2006).
Using PTGui it was done in some minutes - because PGTui set CPs on the strict geometrical window-frames. APG ignored the frames completely and set CPs in the
reflecting clouds . . that can´t work. I didn´t find a way to work around that with APG.
PTGui did very well and very fast here.
But in other cases PTGui can´t do it at all.
In my eyes APG definitely is ahead of PTGui - they both use different strategies, both work, but APG has some advantages.
best, Klaus
Last edited by klausesser (2012-08-11 12:48:07)
Offline
Destiny wrote:
No.. not the case at all.. But I also realise you are set with your opinion so there is no chance of changing it.. But I know you are wrong... I shouldn't be but it is a fact... PTGui and apg interpret the NNP differently.. Same camera, same lens... with different results....
Destiny..
Destiny - i don´t doubt your words. I doubt your implication is correct.
There in only ONE npp - it moves inside the lens when you change the focal-setting . . . and it definitely moves in zoom-lenses changing the focal-length. Here you have to find an npp for each
focal-length.
A different kind of using lens-models can lead to slightly (!) different geometrical results in stitching - that´s what can produce different results in APG and PTGui.
Klaus
Offline
Artisan S. wrote:
Now that would be the case IF and I say IF PTGui is more tollerant of NPP deviations then APG....
PTGui needs correct inputs. APG is much more tolerating non-correct inputs.
"the guy from Germany" . . 
Last edited by klausesser (2012-08-11 12:44:25)
Offline
What I need to do is wait for Alex j to test stitch my images which I achieved Excellent for the quality of stitch and visually too using PTGui.. Whereas the results from apg are no where as good which can also be seen in the nadir footprint. However, if I move the camera/lens by about 5mm, the results are much better but the footprint is still terrible so requires a lot more work to edit.. To be honest, for these images I should just be able to load and stitch since there is nothing to prevent them from stitching perfectly in apg since I am using a motorised head to capture the images... I have also tried my D800 with the 10.5... Using PTGui its just amazing but I cannot say the same using apg, I just do not know why, but I can tell you that if I move the camera/lens, the results are better.. So, does this mean the software respond to different NNP configurations or not.... To me it seems to be the issue... Looking at the image below, it is clear to me that something strange is happening. There is no doubting that there is a difference.. The apg pano has many very small faults in the stitching.. You can see one of them at the tip of the rail, they are small but why is this happening???..
Destiny...
Last edited by Destiny (2012-08-12 01:52:36)
Offline
Artisan S. wrote:
It seems that Autopano uses the SIFT algorithm
Which is widely known since Kolor was founded some 8 (?) years ago . . 
Klaus
Offline
As we can see in the stitching version from APG with this tip of the rail, APG is there using a other image to fill the area.
And in this case, we see a small shift.
So for me it looks like a non perfect nodal point adjustment, that PTGui can better handle. Isn't it ?
Consider that most lenses have a small different nodal point when in short focal distance than at its infinity setting, even if it has internal focusing.
Destiny, which lens you are using? Are there strong distortions to correct?
Offline
lumelix wrote:
Consider that most lenses have a small different nodal point when in short focal distance than at its infinity setting, even if it has internal focusing.
Hi Martin!
Basically that´s correct - but only when you change the focus. Usually you don´t for shooting the Nadir - doesn´t matter when the head gets blurred . . because it´s getting edited anyway.
If you don´t change the focusing for shooting the Nadir the NPP stays where it was.
It´s either the different lens-modules they use or the different algos . . . or both
Guess you also experienced to never get identical results when you do a stitch twice . . guess that´s in the nature of things 
and usually doesn´t matter.
best to you, Klaus
Maybe (!) also there´s a minimal miss-match in adjusting the NPP - which PTGui compensates better than APG . . . who knows
Last edited by klausesser (2012-08-12 17:38:06)
Offline
I agree! I mean exactly this: When change the focus distance.
This happens when focus from infinity to close up
The problem is, who knows this changing NP of his lens and who considered that, when working in confined rooms?
Normally we all use the settings for infinity.
And then we work at a close distance and see bad stitching results
Last edited by lumelix (2012-08-12 17:39:46)
Offline
lumelix wrote:
I agree! I mean exactly this: When change the focus distance.
This happens when focus from infinity to close up
The problem is, who knows this changing NP of his lens and who considered that, when working in confined rooms?
Normally we all use the settings for infinity.
And then we work at a close distance and see bad stitching results
Yeah - but using a fisheye @f8 your DOF is wide enough for not needing to change the focus: a sharp picture of the arm´s base rarely is really essential . .
best, Klaus
Offline
The Nikon 14-24 mm is a other beast in this context.
As a extreme WW it's DOF isn't as deep as with a fisheye.
And as a very complex Zoom lens, it changes the NP too.
So when you work with this lens in small rooms like this caravan above, focus and NP settings are essential
Offline
lumelix wrote:
The Nikon 14-24 mm is a other beast in this context.
As a extreme WW it's DOF isn't as deep as with a fisheye.
And as a very complex Zoom lens, it changes the NP too.
So when you work with this lens in small rooms like this caravan above, focus and NP settings are essential
Yep. I woudn´t use the 14-22 in such a place. They invented fisheyes for that . . .
But even using a 14-22 @f8 you wouldnßt need to focus too much. As is said: no need to have a sharp image of your head´s base - but of
the ground. And here the DOF @f8/11 covers the ground anyway! (forget refraction . . )
best, Klaus
PS.: "problems rise with resolution" . . .
Last edited by klausesser (2012-08-12 19:15:24)
Offline
klausesser wrote:
PS.: "problems rise with resolution" . . .
Yep: D800 = 36 MP
That's what I thought when asking about the lens Destiny has used there.
DOF definition and airy disc theory is a little bit tricky with this sensor.
Last edited by lumelix (2012-08-12 19:48:03)
Offline
lumelix wrote:
DOF definition and airy disc theory is a little bit tricky with this sensor.
You bet!
best, Klaus
Offline
Hey... Well, this nadir images here are from using my D90 and 10.5 fisheye... If I move the NPP settings to accommodate apg better, the rail still juts out but the stitching is better.. So, the question I raised, does the NPP change according to the software used.. Perhaps not but software certainly interprets stitching the images differently...
So then I tried out my D800 and put my 10.5 fisheye on which automatically crops to dx... I set my NPP and in PTGui I achieved "Excellent" as a result and visually beautiful too, faultless stitching and the nadir footprint is perfect looking.. I put the same images into apg with very similar results when using my D90, poor stitching and yuck nadir image.. I did not try to change the NPP settings since I assumed it would be the same results as using my D90 with 10.5... I can get much better results in using apg but I would have to change my NPP to achieve it but then the results from PTGui would be very different... It might even say, poor... So, for me, its not a matter of wanting to use PTGui, its a matter of it works with my gear much better than apg... Which is a shame since apg is a much more powerful software app with a lot more controls.. but I do not want to spend hours playing with control points when I should be able to achieve acceptable results especially since I am using a VR Drive II pano head which captures images very accurately...
I had a setting of f8 with an ISO of 200 with 3 bracketed shots using my D90 and 10.5 when I captured these caravans.. I then fused the images using Photomatix. Perhaps there is a Photomatix thing with apg, perhaps it has missing data that apg gets confused over.. We already know it has issues in determining the lens type... So perhaps this is the issues..
A point was raised above where you get different results from the same images, same software on different days... This is true so why is that.... I have put images into apg and PTGui and not been happy.. Thinking it was my settings I tried again a day or two later with the same image but with different results..
One thing is certain, if the NPP is set to apg perfectly, the detection is amazingly quick and the stitching/rendering is also very fast.. This is a good indicator to the final results.. If very slow, you might experience stitching issues...
Destiny..
Last edited by Destiny (2012-08-15 00:23:53)
Offline
Hi Destiny
Very interesting case.
It's clear, the NP insn't dependent on the software you are using. It's dependent on the camera/lens set and (for a really small part) the focal settings.
But as you described, the results are not always the same and differs between APG and PTGui.
The results of lens distortion correction are dependent from the lens/Sensor informations.
So when you are using your 10.5 FE (DX) on the D800 (FX), check that there are the correct settings in APG for both, lens and Sensor.
But I'm wondering if this combination DX lens / Fx sensor is fully supported in the distortion algorithm of APG ?
Offline
Jean-Phi38 wrote:
i spend some few days trying to perfect my NPP (no Parallax Point) and tried different settings with my gear (Canon 7D + Sigma 10-20 @10mm + NN4) the Lower Rail settings is 54mm and Upper Rail 100mm.
when doing a 360*180° (with 16 imgs; 4*-60°, 8*0°, 4*60°) i have some shifting on the ceiling.
I send the images to Nodal ninja forum, and a guy with PTGui stitched my panorama without problem and no shifting.
Hello,
I had simiar issues with my Tokina 11-16 @ 11mm ultra wide angle lens. Did you try to set the distortion model to 2nd order + shift? See the video tutorial here: http://www.autopano.net/wiki-en/action/ … correction
Regards,
Jens
Offline
New day new test just to see.. So I put the settings to "force difference" and "2nd Order", but I have already done all this before and every other combination.. Not much point as you can see since its even worse, also lots of small stitching faults.. So I will stick with PTGui since all I have to do is load the images and press stitch.. Sorry..
Destiny..
Offline
Hi Ed.. Yes, It would be interesting.. But I am guessing kolor are too busy since I have already sent my images to them about two months ago to Alex J for testing as you can read in the other thread mentioned at the beginning of this thread... No response so far.. So probably a bit busy.. I would gladly use apg if it worked for me.. But I just do not know what settings I am best using.. The strange thing is.. the preview does not show the nadir as seen in the rendered output..
Destiny..
Offline
There is definitely a huge difference between APG and PTGui Pro when it comes to stitching fish-eye images.
PT Gui stitched this without any manual intervention (apart from setting the image circle crop) in about a minute:
http://08.ae/panos/downtownsunset/sunset.html
That's 3 shots from a Canon 8-15 on a Phase One IQ180.
This is the starting point with APG:
I can't help thinking the two programs are optimised for different things. APG is fantastic at large panos. It will do in hours what PT Gui takes weeks. Literally.
Offline
Hi... apg is a much more powerful software application than PTGui, it is ideally suited for giga pixel panos and seems to do a great job.. I personally find it to be fine for stitching a normal single layered shot pano sphere on most occasions using my 10.5 fisheye with D90. It can also cope with poorly captured images better than PTGui... For me it most defiantly the stitching of bracketed shots that lets apg down.. The LDR and HDR are both terrible... Yet most pano photographers capture using bracketed shots from 3 to 9+.. The apg software is getting better but I feel it still has a long way to go with bracketed shots...
Since apg's own LDR and HDR output are not up to my needs, it frustrates me that my fused images from Photomatix seem to cause issues with apg.... So for now I have given up trying...
Destiny..
Offline
Pages: 1 2
