Ad
Ad
Ad
Pages: « 1 [2] 3 »   Bottom of Page
Print
Author Topic: Confused about Panorama Setups (panorama clamp vs. leveling base)  (Read 4158 times)
Tony Jay
Sr. Member
****
Offline Offline

Posts: 2028


« Reply #20 on: January 09, 2013, 05:05:58 AM »
ReplyReply

...Several posters were speaking of what the stitching software can do. That's great info and the improvements nowadays are awesome, but that's not what i was driving at. Using the software more heavily may indeed, in the end, be the more practical solution, but I was more looking for what the "ideal" route was for getting as much correct at the time of the shot is...
Believe me when I say this: trying to stitch poor images into a panorama must be the most frustrating and unedifying experience there is in post-processing.
With the best possible series of images you have a chance, with the best software, of getting a good result, but there is no guarantee.

So, to get any sort of result requires maximum leverage of your shooting ability and maximum facility with your stitching software.

Tony Jay
« Last Edit: January 09, 2013, 05:08:03 AM by Tony Jay » Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3007


« Reply #21 on: January 09, 2013, 05:44:38 AM »
ReplyReply

Thanks for all the recent thoughts. I don't have the time to go over all of them right now, but I was thinking about this last night and think I understand what the issue is now.

Here's an extreme example. Take the camera, mounted on a tripod/ballhead and flip it over 90 degrees and aim it at the ground. Rotate it with a pano clamp and it's going to spin around to include the sky. Rotate it with the pano base of the ballhead and you'll scan the ground, which was most likely your objective.

I doubt it. When you spin the camera around with the pano base of the ball head, each image will have a different position of the entrance pupil of the lens in a circular path around the tripod center, which makes it impossible to avoid parallax between the image tiles, which in turn creates problems with seamlessly overlapping stitching of 3-dimensional surroundings.

Try adding a fore/aft sliding bar to your top clamp setup to eliminate the entrance pupil parallax in a single row, and you'll see the benefit of using the top clamp rotation. Only with this latter approach will overlapping tiles have zero parallax in the overlap zone, which makes reliable automatic stitching possible.

Quote
I'm guessing this is why the people advocating that a pano base, with a leveler if you so choose, is going to come closer to achieving the results you're looking for in more varied situations, ie: tilting the camera up or down.

Remember, the usual goal is being able and stitch the adjacent pano tiles, without parallax issues in the overlap zones spoiling the fun. The goal is usually not just to simply cover a given Field of View with lots of problems to make a coherent composite image out of it. If that were the objective, then hand-held shooting would suffice. Especially with demanding foreground features and predictable patterns, eliminating parallax is the key to success.

Funny enough, there is also a possible advanced use for parallax on flat surfaces, such as your thought experiment with the camera pointing down and spinning round the ballhead base, but that requires stitching software that can cope with camera position offsets and masking. With some math and masking, it is possible to eliminate the tripod (legs) from showing in the so called 'Nadir image', e.g. in a 360 degree VR pano.

Cheers,
Bart
Logged
Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1628



WWW
« Reply #22 on: January 09, 2013, 07:44:15 AM »
ReplyReply

I hope this isn't a dig at Ellis.

Tony Jay
Of course it is. But I don't try to write for photography related publications - I've been making good money doing it for nearly ten years.
Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
aman74
Newbie
*
Offline Offline

Posts: 11


« Reply #23 on: January 09, 2013, 07:59:06 AM »
ReplyReply

I doubt it. When you spin the camera around with the pano base of the ball head, each image will have a different position of the entrance pupil of the lens in a circular path around the tripod center, which makes it impossible to avoid parallax between the image tiles, which in turn creates problems with seamlessly overlapping stitching of 3-dimensional surroundings.

Try adding a fore/aft sliding bar to your top clamp setup to eliminate the entrance pupil parallax in a single row, and you'll see the benefit of using the top clamp rotation. Only with this latter approach will overlapping tiles have zero parallax in the overlap zone, which makes reliable automatic stitching possible.



Found this thread:

http://photo.net/nature-photography-forum/00OE1U?start=0

From a few posts down:

"The reason to use a leveling base for shooting stitched panos would be if you wanted to create a pano where the camera was not pointed directly at the horizon (level). For example, a pano looking downward into a canyon or pano looking slightly upward at mountains.

If you use a panning clamp (like RRS's) and tilted the camera upward the arc of the pan would curve downward the farther off center you got. If you tilted downward the inverse is true, edges of the photo curve up.

However, if you use a leveling base combined with a ballhead with built in panning base (not a panning clamp), then you can level the base of the ballhead with the leveling base and then pan left right on a perfectly horizontal axis using the ballhead's panning base. Then even if you tilt the camera down or up with the ballhead, the axis of rotation remains horizontal.

Hard to explain, but easy to see once you try it out."

That's what I'm getting at...
Logged
simonstucki
Jr. Member
**
Offline Offline

Posts: 64


WWW
« Reply #24 on: January 09, 2013, 08:52:33 AM »
ReplyReply


If you use a panning clamp (like RRS's) and tilted the camera upward the arc of the pan would curve downward the farther off center you got. If you tilted downward the inverse is true, edges of the photo curve up.

but this way you don't get any parallax problems if you use a "nodal" slide/rail. however at some point you might not get the image you want, simply because there is too  much of a curve when you stitch the images.

However, if you use a leveling base combined with a ballhead with built in panning base (not a panning clamp), then you can level the base of the ballhead with the leveling base and then pan left right on a perfectly horizontal axis using the ballhead's panning base. Then even if you tilt the camera down or up with the ballhead, the axis of rotation remains horizontal.

yes but then you get parallax errors because you move the entrance pupil of the lens. with some subjects (distant mountains) it might work perfectly but not so much with an important foreground. to avoid that you need to use a 3d pano head.
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3007


« Reply #25 on: January 09, 2013, 08:56:49 AM »
ReplyReply

"The reason to use a leveling base for shooting stitched panos would be if you wanted to create a pano where the camera was not pointed directly at the horizon (level). For example, a pano looking downward into a canyon or pano looking slightly upward at mountains."

Yes, and I hope that it's well understood by now that that won't affect the horizon's curvature or waviness, provided that the correct Pitch angle is used in the stitching software. As was explained (or at least hinted at) in that thread, leveling the base of the Ballhead and then allowing a joint above that with a clamp which destroys the leveling (as far as the camera is concerned) for looking down or up, doesn't help to reduce parallax. It only helps to reduce a curved image path.

That above quote only makes sense with a complete multi-row pano setup, with the entrance pupil aligned for all 3 axes, Yaw / Pitch/ Roll. With a simpler, but often good enough single row setup, there is no way to avoid both a curved image path and parallax at the same time on a 3-dimensional scene. Only with clever software and a flat plane surface can we remove parallax after the fact.

Cheers,
Bart
« Last Edit: January 09, 2013, 08:59:53 AM by BartvanderWolf » Logged
aman74
Newbie
*
Offline Offline

Posts: 11


« Reply #26 on: January 09, 2013, 09:10:39 AM »
ReplyReply

but this way you don't get any parallax problems if you use a "nodal" slide/rail. however at some point you might not get the image you want, simply because there is too  much of a curve when you stitch the images.

Couldn't you use the nodal slide with the ballhead panning though?

Quote
yes but then you get parallax errors because you move the entrance pupil of the lens. with some subjects (distant mountains) it might work perfectly but not so much with an important foreground. to avoid that you need to use a 3d pano head.

Again, can't we use the nodal slide for this?
Logged
aman74
Newbie
*
Offline Offline

Posts: 11


« Reply #27 on: January 09, 2013, 09:20:31 AM »
ReplyReply

Yes, and I hope that it's well understood by now that that won't affect the horizon's curvature or waviness, provided that the correct Pitch angle is used in the stitching software. As was explained (or at least hinted at) in that thread, leveling the base of the Ballhead and then allowing a joint above that with a clamp which destroys the leveling (as far as the camera is concerned) for looking down or up, doesn't help to reduce parallax. It only helps to reduce a curved image path.

We can't use the nodal slide for this? I'm assuming not or you guys would have mentioned it.

Quote
That above quote only makes sense with a complete multi-row pano setup, with the entrance pupil aligned for all 3 axes, Yaw / Pitch/ Roll. With a simpler, but often good enough single row setup, there is no way to avoid both a curved image path and parallax at the same time on a 3-dimensional scene. Only with clever software and a flat plane surface can we remove parallax after the fact.

Cheers,
Bart

Ok, I'll have to re-read that thread and see what was hinted at.

I think I'm starting to get what you're driving at. Lemme see if I have this right. If we wish to stay with a single-row setup, the panning clamp makes more sense as the curved image path is easier to deal with than parallax errors?

Sorry that I'm having a hard time getting all this straight.
Logged
Slobodan Blagojevic
Sr. Member
****
Online Online

Posts: 4998



WWW
« Reply #28 on: January 09, 2013, 09:38:00 AM »
ReplyReply

...Sorry that I'm having a hard time getting all this straight.

And the hard times will continue indefinitely until you actually start doing it yourself. Otherwise, it will be like trying to become a brain surgeon by reading books, or, worse, Internet debates about it.
Logged

Slobodan

Flickr
500px
Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1628



WWW
« Reply #29 on: January 09, 2013, 11:28:45 AM »
ReplyReply

+1 to Slobodan! Trying to hold in your head how all of this works is very very difficult. Once you start, even without specialized panoramic stitching gear, will give you insight into the mechanics of shooting stitched panoramas
Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1628



WWW
« Reply #30 on: January 09, 2013, 11:40:51 AM »
ReplyReply

Attached is a single row panorama with the camera level and horizontally rotated at the camera platform. The long side of the camera's sensor was oriented vertically. No pitch or other corrections were done in the stitching software. Cylindrical Projection
Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3007


« Reply #31 on: January 09, 2013, 12:45:28 PM »
ReplyReply

Sorry that I'm having a hard time getting all this straight.

Hi,

No problem. It creates problems to envision the consequences for a lot of people, until they understand/embrace the concept of avoiding parallax as the technical goal. Actually doing it helps when that goal is kept in mind. Some subject matter may be somewhat forgiving and with proper software blending we may cover up some shortcomings, but the best chance of success comes with proper technique.

To avoid entrance pupil parallax for single row panoramas (whether tilted up, down, or on its side plays no role) we need to satisfy 2 alignments for the intersection of orthogonal rotational axes at the same time in a single point, and for multiple-row panoramas that would require alignment of 3 rotational axes in a single intersection point.

In the single row pano case, we need to make sure that the camera's Yaw axis and the optical axis (Roll axis) intersect at the entrance pupil position. That requires a lateral shift (only if the tripod socket is not aligned with the optical axis), and a fore/aft 'nodal' or rather No-Parallax Point shift, usually backwards (the amount varies with lens focal length, and perhaps a bit with focus distance). The best way to achieve this, is with a rotating clamp positioned directly under the optical axis (no lateral offset), and a fore/aft sliding bar to move the entrance pupil of the lens directly above the rotating clamp.

With that done correctly, we can happily rotate around the camera's Yaw axis, without introducing any parallax, and stitching will be easy, even with difficult subject patterns. It will not prevent a curved path of image tiles relative to a horizon when we tilt the clamp plane to a non-leveled angle, but straight lines (including the horizon) will remain straight (not level, but straight) when a rectilinear projection is used for rectilinear lenses and the correct Pitch value is set in the stitching software. When the camera is used in portrait orientation, the vertical FOV is often sufficient enough to avoid the need for a heavier, more complex, multi-row pano assembly.

When we relax our No-Parallax Point (NPP) requirement, everything becomes much harder to understand, and to stitch.

Cheers,
Bart
« Last Edit: January 09, 2013, 05:27:56 PM by BartvanderWolf » Logged
Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1628



WWW
« Reply #32 on: January 09, 2013, 01:12:48 PM »
ReplyReply

Bart's explanation and analysis, is spot on.

As Bart points out, even with the camera carefully levelled,  life is far easier with a rotating clamp located as close as possible to the entrance pupil (AKA"nodal point") rather then separated from it by the body of the tripod head.

Using a rotating clamp at the camera platform level isolates  the camera's rotational plane from the pitch (fore/aft) and roll (lateral) angles you've set with the heads controls.
« Last Edit: January 10, 2013, 04:27:11 PM by Ellis Vener » Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
simonstucki
Jr. Member
**
Offline Offline

Posts: 64


WWW
« Reply #33 on: January 09, 2013, 05:54:06 PM »
ReplyReply

Couldn't you use the nodal slide with the ballhead panning though?

Again, can't we use the nodal slide for this?

hmmmmmm actually I think that might work "quite well" (compared to multi row) for a single row tilted panoramas. of course you would have to reset the nodal rail when you change the tilt. but you could find an equation to help you set the nodal rail for any different tilt. it is going to be slightly different for different lenses/cameras. so theoretically it should work as well for multi row panos, but then you would have to reset the nodal rail for each row. I think that's when I would buy a real 3d pano head, actually I would buy one way earlier, because you can save the leveling base for the ballhead (or the ballhead vice versa) since you only have to level the panning base for the 3d head so you probably don't even save that much weight.

am I right or have I lost it completely?
Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 6897


WWW
« Reply #34 on: January 10, 2013, 05:05:39 AM »
ReplyReply

+1 to both!

Erik

+1 to Slobodan! Trying to hold in your head how all of this works is very very difficult. Once you start, even without specialized panoramic stitching gear, will give you insight into the mechanics of shooting stitched panoramas
Logged

Ellis Vener
Sr. Member
****
Offline Offline

Posts: 1628



WWW
« Reply #35 on: January 11, 2013, 02:16:01 PM »
ReplyReply

I have attempted a drawing to show why the geometry of using a rotating clamp on the head's camera platform works better. I have kept it simple and am only considered pitch (for/aft) tilt angle and not also roll (lateral) tilt angles.

While you could adjust the position of the camera forward or backward to keep it aligned with the rotational axis of a base rotator  you'd have to change that position anytime to maintain parallax correction you used a different tilt angle to maintain alignment. That to me is more of a pain than just buying a rotating clamp and a slide.

If I have this wrong please correct.

two final notes:

My gut feeling is that while programs like PTGui, AutoPano Pro and the PhotoMerge script in Photoshop CS6 (although not in earlier versions of Photoshop) have gotten very good at handling parallax errors ,  for best results minimizing how much interpolation the stitching program must do to correct for parallax errors is a better way to work yielding fewer problems that have to corrected by hand later.

But if the final panoramic will be reproduced at a small size and/or  the differences in the near/far relationships in the photo approach zero  those kinds of errors may not matter. 
Logged

Ellis Vener
http://www.ellisvener.com
Creating photographs for advertising, corporate and industrial clients since 1984.
Jason Denning
Full Member
***
Offline Offline

Posts: 228



WWW
« Reply #36 on: January 11, 2013, 03:43:24 PM »
ReplyReply

You forgot to list the 617 camera as an option.
Logged

www.jasondenning.co.uk


Fotoman 617 with Rodenstock 55mm, 90mm and 180mm lenses
Mamiya 645 Pro TL, and every lens mamiya made.
Sony A7 with 35mm and 55mm Primes
OldRoy
Sr. Member
****
Offline Offline

Posts: 404


WWW
« Reply #37 on: January 11, 2013, 05:22:29 PM »
ReplyReply

If this thread was the first thing I'd come across about pano stitching I think I'd have taken up golf instead.

You need a pano head and the software. I suppose you can do without the pano head under certain limited circumstances. You need to set up the pano head so that rotation in both axes is about the no-parallax point; there are plenty of tutorials that show you how to do this. Here's a good source:
http://www.johnhpanos.com/tuts.htm

Learning PTGui (and, I guess, AutoPano) is a bit of a challenge but it's the key to accomplishing the stitching/blending sucessfully.

Amazing how the topic of levelling before shooting comes up time after time - a total red herring.
Roy
Logged
DennisG
Newbie
*
Offline Offline

Posts: 5


« Reply #38 on: January 12, 2013, 12:39:25 PM »
ReplyReply

Concerning Ellis Vener's posted pano of the railroad tracks:  it does not seem to me that there is a camera pitch, or post-processing pitch adjustment that will make both the tracks and the horizon undistortedly linear (i.e. straight, uncurved), from his shooting position.  Only a QTVR file will do that, which looks very natural but is not printable, or a rectilinear projection, which seems to give poor results, at least in PTGui.  If there is a solution to this problem, please let me know.

Thanks
Logged
simonstucki
Jr. Member
**
Offline Offline

Posts: 64


WWW
« Reply #39 on: January 13, 2013, 02:16:36 PM »
ReplyReply

Concerning Ellis Vener's posted pano of the railroad tracks:  it does not seem to me that there is a camera pitch, or post-processing pitch adjustment that will make both the tracks and the horizon undistortedly linear (i.e. straight, uncurved), from his shooting position.  Only a QTVR file will do that, which looks very natural but is not printable, or a rectilinear projection, which seems to give poor results, at least in PTGui.  If there is a solution to this problem, please let me know.

Thanks

poor results in what way? (concerning the rectilinear projection)

and of course a rectilinear projection doesn't work with a 180 angle of view. the wider you go, the more extreme gets the "rectilinear wide angle effect" (sorry don't know how to describe it, just look at some pictures with a 14mm on a full frame).
also the wider you go (when you choose rectilinear projection), the more have the images on the edge of the stitched image to be stretched. so you loose resolution there. the solution to that problem is to start with a higher center resolution (much more than you need/want) so that you still have the desired corner resolution (of course when you then see the amazing center resolution, then you might also want this in the corner and you might want choose an even longer focal length, and then again and again and before you go completely crazy just keep in mind that you probably already have a much higher resolution than if you shot this image in one shot, so just ask yourself how large do I wan't to print it and then stop:).
Logged
Pages: « 1 [2] 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad