Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
2 X 180 nadir tripod shots
Posted: 19 May 2008 at 17:02 GMT
|
|
Today I shot another pano with the intention of generating a stitchable nadir using the 2 X 180 -90 deg tripod shots method. Previously I have always omitted this stage and patched the nadir (after creating cubefaces) using a hand held shot. To aid this I usually distribute some homemade targets around the tripod - which helps greatly in transforming the patch to fit.
Call me a dimwit but I am - yet again - completely unable to get two tripod-shot nadirs to successfully fuse. There is a huge component of error between the two shots which makes alignment effectively impossible.
I have once successfully done this, recreating a couple of previously completed panos using tripod shots which I hadn't used before, however it required the targets to aid alignment and some distortion to get a fit.
I feel like an idiot as this method is obviously widely used and yet there is clearly a huge error in the alignment of my 2 X 180 shots which is almost impossible to reconcile. At some point in the fused image there is always extremely visible misalignment.
I have asked this question more than once, but I ask it again in the hope that someone can explain why it is occurring. I assume that my entrance pupil alignment is adequate as I am not getting parallax errors. It seems to me that it is probably the fault of my "Leaning Tower" NN3 - a suggestion that nobody has made so far.
|
|
alert moderator
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: 2 X 180 nadir tripod shots
Posted: 19 May 2008 at 17:49 GMT
|
|
Roy, Try my draft tutorial which I knocked up this morning with you in mind.
www.johnhpanos.com/ptgvpt.htm
(The two nadirs on the tripod are actually at 90 degrees rather than 180, but the method is just the same).
John
|
|
alert moderator
|
|
Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: 2 X 180 nadir tripod shots
Posted: 19 May 2008 at 18:16 GMT
updated: 19 May 2008 at 18:17 GMT
|
|
Thanks John
I've printed the tutorial and downloaded the sample images. This is a slightly different approach to the layer mask method that I've been trying (mostly unsuccessfully). It introduces a few concepts that I wasn't aware of.
I will give it a try and let you know how I get on.
Roy
|
|
alert moderator
|
|
Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: 2 X 180 nadir tripod shots
Posted: 20 May 2008 at 17:48 GMT
updated: 20 May 2008 at 17:56 GMT
|
|
John
this technique is a revelation (well, to me it is anyway!) I actually took the liberty of trying this for the first time using some shots of my own that I made for the purpose - keeping your test shots as a fallback for any failure; I didn't need to use them. I haven't corrected any of the control points so the stitches are a bit out in a few places however the results are really excellent for a first attempt by this method. The tutorial is lucid and unambiguous. However I have a few questions.
1) I am using a template derived from a nadir-less 7-shot sequence. I applied this at the usual stage (ie immediately after loading the image set) which I thought might well be a source of error. What do you suggest?
2) This test pano I shot @ -5 deg pitch. Church ceilings are pretty gloomy even with an HDR set. From your own experience, do you feel that -10 deg is liable to give consistently useable zenith stitching?
3) The test location has a carpeted floor. As a fallback, rather than risking having to go back, I scattered some targets on the otherwise featureless nadir region around the tripod. This necessitates dismantling the final equi into cubefaces (Pano2QTVR, Lanczos 3) editing in PS and reversing the procedure. Apart from the time required, do you feel that this step should be avoided if possible. Put another way, can I afford to abandon the safety net afforded by the "targets"? How would you usually deal with this problem?
4) Does the optimisation sequence - first FOV plus H & V shift, then re-optimise adding 3 distortion parameters - remain the same? I know this is a dumb question but I still don't have a clear picture (to understate the case, and despite reading your optimisation tutorial) of how the optimiser processes the various parameters.
5) The first pass through the final part of the procedure actually gave me slightly worse optimisation of the HH patch after checking the viewpoint correction box! Is this indicative aof anything I should bear in mind?
I will now try creating a couple of panos for as accurate as possible results. Thanks again for this extraordinarily useful tutorial. I'll add any further queries or observations to this thread.
Roy
|
|
alert moderator
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: 2 X 180 nadir tripod shots
Posted: 20 May 2008 at 20:15 GMT
|
|
Roy, I was really pleased to hear that you managed to cope so well with the tutorial.
Skridlov said: 1) I am using a template derived from a nadir-less 7-shot sequence. I applied this at the usual stage (ie immediately after loading the image set) which I thought might well be a source of error. What do you suggest?
It's probably ok. I would generally recommend generating a template with a nadir and zenith taken on the pano head as this helps to get an accurate fov.
2) This test pano I shot @ -5 deg pitch. Church ceilings are pretty gloomy even with an HDR set. From your own experience, do you feel that -10 deg is liable to give consistently useable zenith stitching?
Again, this should work ok, but it may help to take the zenith shot at +70 pitch and orient the camera so that the lower side of the image captures brighter detail below the ceiling.
3) The test location has a carpeted floor. As a fallback, rather than risking having to go back, I scattered some targets on the otherwise featureless nadir region around the tripod. This necessitates dismantling the final equi into cubefaces (Pano2QTVR, Lanczos 3) editing in PS and reversing the procedure. Apart from the time required, do you feel that this step should be avoided if possible. Put another way, can I afford to abandon the safety net afforded by the "targets"? How would you usually deal with this problem?
If the carpet is featureless, then it is a good idea to add some carefully placed targets. These can be quite small and you don't need many. I've never generated cube faces myself for editing purposes as I prefer PTEditor. This seems to do the least damage to the panorama, but you should be quite safe enough editing with Pano2QTVR as you describe.
4) Does the optimisation sequence - first FOV plus H & V shift, then re-optimise adding 3 distortion parameters - remain the same? I know this is a dumb question but I still don't have a clear picture (to understate the case, and despite reading your optimisation tutorial) of how the optimiser processes the various parameters.
There are no rules. You optimize those parameters that need optimizing. If you have a calibrated set of lens parameters, there should be no need to optimize them. However, if you are finding it difficult to get a good optimization, then provided there are sufficient control points well spread out, then you might find a useful improvement by optimizing them. Things don't always go exactly to plan, however much care you take.
5) The first pass through the final part of the procedure actually gave me slightly worse optimisation of the HH patch after checking the viewpoint correction box! Is this indicative aof anything I should bear in mind?
I haven't done a large number of viewpoint optimizations. In my experience, it either works very well or badly. In the latter case, trying different positions of the nadir by manual adjustment usually encourages the optimization to work after a few attempts. (Though this was before I found that rotating the panorama up by 90 degrees made the whole process easier and more reliable. Hans made the same discovery). If you can't get a good result, I would check and maybe change the control points if possible. I've so far always managed to get a very good result without much trouble.
John
|
|
alert moderator
|
|
Pete Loud
Posts: 152
Location: United Kingdom
Registered: 14 Oct 2006
|
|
Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: 2 X 180 nadir tripod shots
Posted: 20 May 2008 at 21:50 GMT
|
|
Hi again John
well, there are some interesting wrinkles emerging. The sequence is absolutely critical here it seems. One mistake I made was to forget to optimise after setting the T1, T2 levelling cp's - but before adding the HH patch shot. I couldn't find any way of backing out of this situation. I tried deselecting the HH parameter check-boxes and re-optimising but it wouldn't have it. The alignment dialog wouldn't have it either - but I assume it uses the same code anyway.
In the end I removed the HH shot, optimised, re-added it and proceeded according to your tutorial. I still had some odd complaints when I came to generate the panorama that I hadn't optimised even though I had definitely done so.
I'm not entirely happy with the levelling on this test, however. Is it possible, without dumping the HH patch, to re-level?
I need to do a couple more to get a better feel for the technique but I'm delighted with my first attempt. I would never have got to this without your tutorial (it goes without saying) not least because I thought that "viewpoint correction" was designed for use with a single oblique nadir shot.
Of course there will be more questions!
Many thanks John.
|
|
alert moderator
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: 2 X 180 nadir tripod shots
Posted: 20 May 2008 at 22:24 GMT
|
|
Roy, It's best to leave the levelling until the very end. You can then add the t1/t2 points and either use the optimizer or just use Edit->Level Panorama in the Panorama Editor window.
John
|
|
alert moderator
|
|
Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: 2 X 180 nadir tripod shots
Posted: 20 May 2008 at 23:20 GMT
|
|
Ok John, but this was the problem I encountered - the levelling was inoperative. By now I've been round the houses so many times that I can't recall the exact sequence but there seems to be some problem associated with using the PTGUI rather than Panotools optimiser with this method.
But I am a bit confounded by something quite unrelated which is preventing me from getting a final result. I'm back to the "bad old days" in that I simply can't get the basic seven images to stitch properly. I've established this by creating a new project and just stitching these without the additional complication of the nadirs (although the result's the same in both cases). There is a stitching error below the equator in 5/0 and 0/1 which no amount of cp deleting and resetting will solve. There are no inordinately large cp errors overall (it's a bit confusing to use PTGUI's optimiser which has different opinions to the PT version...) but I simply can't get these to stitch. Usually any errors are located in the tropics, and easily rectified.
One change in this set is that I am using -5deg for the 6 round. I hadn't set this value in the image parameters initially. If I set it before running the auto alignment the image orientations become scrambled. If I run the auto alignment, and then subsequently set the pitch to -5 deg the errors are relocated to 2/3 and 3/4 in roughly the same pitch elevation as the previous errors.
In either case do you have any idea why this would occur given the roughly same setup and template that I have used successfully for 15 or more previous panos?
Apologies for thread creep.
Roy
|
|
alert moderator
|
|
michael medina
Posts: 297
Location: portland, oregon, United States
Registered: 27 Jan 2008
|
Re: 2 X 180 nadir tripod shots
Posted: 21 May 2008 at 0:35 GMT
|
|
project files help a lot in this case. i've never used a template and have never had to apply one. you should be fine.
but if you post a project file, you could have all your questions answered quite quickly instead of having john (and others) just guessing what is going on. i'm guessing that this could also improve your opinion of the rotator on the nn3.
so please make up a few project archives so we can take a look at what you are doing
|
|
alert moderator
|
|
John Houghton
Posts: 2317
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: 2 X 180 nadir tripod shots
Posted: 21 May 2008 at 7:35 GMT
|
|
Skridlov said: One change in this set is that I am using -5deg for the 6 round. I hadn't set this value in the image parameters initially. If I set it before running the auto alignment the image orientations become scrambled. If I run the auto alignment, and then subsequently set the pitch to -5 deg the errors are relocated to 2/3 and 3/4 in roughly the same pitch elevation as the previous errors.
Roy, I don't know why you are running auto alignment. You have a template with decent lens parameters, albeit for images not taken at -5 degrees. Start by applying the template. Then set the pitch of the 6 around to -5. Generate control points with the Generate Control Points option on the Control Points menu. Optimize, including fov and the shift parameters but not a,b,c. Examine/correct/delete any bad points. Generally, don't delete bad points that are accurately positioned on solid features. Repeat the optimization and checks until you get a good report and bad points have been eliminated. If you have enough control points well spread out, you can try optimizing all the lens parameters too.
I'm not sure what you mean by "the image orientations become scrambled". Are they scrambled in the output area or the Project Assistant panel or both? The images can be in assorted orientations in the Project Assistant yet be perfectly aligned - but not levelled - in the output area. Levelling sorts that out.
But Michael is right: diagnosing problems is far easier with some hard evidence to play with.
John
|
|
alert moderator
|
|
Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: 2 X 180 nadir tripod shots
Posted: 28 May 2008 at 13:32 GMT
|
|
If the rotator on the NN3 is so good, why is there a replacement?
This is a seperate issue!
|
|
alert moderator
|
|
Smooth
Posts: 1472
Location: Australia
Registered: 21 Jul 2004
|
Re: 2 X 180 nadir tripod shots
Posted: 28 May 2008 at 13:51 GMT
|
|
Roy,
The rotator on the original NN3 works fine (or at least for me and plenty of others) The only issue I ever had with it was if you used it in an anti-clockwise manner. If you did/do the tension knob would undo. This bought about the new improved rotator that comes standard with the NN3 MKII and offered as an upgrade to all owners of older modal Nodal Ninja panoheads for minimal charge. I can tell you these are a far superior design and anyone who has one will tell you "get one" seriously!
Reading your posts over the past months makes me think your issues with stitching are not really anything to do with the panohead but your lack of understanding the PTGui program or something else in your shooting and processing methods. People take hand held panoramas and successfully stitch these images. So even if the panohead isn't adjusted correctly it should pose severe problems.
I also thought that I read that Nick from Nodal Ninja/Fanotec offered you a full refund if you wanted to return it?
Regards, Smooth 
www.smooth360.com
|
|
alert moderator
|
|
Skridlov
Posts: 158
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: 2 X 180 nadir tripod shots
Posted: 28 May 2008 at 17:38 GMT
|
|
There are two issues conflated here - but NOT by me. The NN3 rotator mechanism has nothing to do with this thread at all! But, since others wish to combine them, I'll respond in the hope that it clarifies the subject(s):
The original NN3 rotator has been replaced by another unit - someone please correct me if I'm wrong about this. Is this because the original is so well designed it has to be replaced by something LESS effective? This replacement is chargeable, and for that reason I think it's not unreasonable to comment.
My complaints about the NN3 - significant but not overwhelmingly so - are that
(a) It won't support a D200 without the weight torqueing the vertical post by a few degrees of roll. I admit to being a dummy, but I am well aware that this can be offset in PTGUI by levelling, having done it several dozen times. I just wish it wasn't so.
(b) The NN3 rotator tensioning mechanism is poor. It requires frequent adjustment and if set to eliminate slop it's very easy to overtighten. The locking screw continually slackens off - always if I rotate anti-clockwise (& yes I've read the instructions). I would maintain that it is a bad design. I speak as someone who has also created and sold deficient products (to Hitachi amongst others) which I had to subsequently redesign. I frequently had to bear the cost of modifications - and refunds.
The NN3 is a fairly simple device - with a few faults in the case of the model I own - at what is a pretty steep price in the UK. I have the right to an opinion about this - which seems to be unacceptable here. I also think that it's nicely presented, compact etc etc etc. Which I've also stated more than once. Why this should be so controversial I don't understand.
A full refund is a fair offer, but I'd prefer a free replacement of the faulty rotator mechanism. Again, if it ain't broke, why the fix? I need a panhead of some kind and the alternatives are outside my price range so I have little choice but to continue to use it. I'm not going to enter into any more dialogue about this.
Problems using PTGUI are a totally separate issue. To put it mildly, it is not an easy program too use. I'm over sixty and no longer as quick as I once was - not that I was particularly clever to begin with.
After writing the above I no longer have the energy to write about the original subject of the thread. I'm weary of all this. Maybe I should find a different way of wasting my time.
|
|
alert moderator
|
|
Hans Nyberg
Posts: 866
Location: Denmark
Registered: 28 Aug 2005
|
Re: 2 X 180 nadir tripod shots
Posted: 28 May 2008 at 18:27 GMT
|
|
I'm 67 and I only have 9 years basic school.
I hate advanced math and my eyes needs a +4 if I want to use my MacBook Pro 17.
But I never had any problems stitching with PTGui even if I sometimes use the lenses at nodalpoints 5 mm from the perfect spot.
I never take nadir with the camera at -90 degree as it is waste of time.
Always -10 degrees with a fullframe fisheye and a nadir with tilted tripod.
NodalNinja 3 has no problem even with my Canon 5D but as it is a little to small for the zenith I removed the rotator and use my Manfrotto 300N for it.
I prefer this setup instead of the NN5
Hans
|
|
alert moderator
|
|
Prev
