![]() |
|
|
|
|
|
||||||||||
|
| User list | Rules | You are not logged in.
Hi All
I am doing a few vertical panoramas at the moment and find that the results I am getting are not up to the usual APP standards. I have found APP lacking in its support for this type of image.
The software tends to build the panorama laid out at 90 degrees to the correct orientation. I have no issues with seeing my work sideways, but in circular projection, which is mostly what I use, I am then unable to use the vertical lines tool. If I tell APP to rotate the image to its correct orientation, it distorts the image. If I then use the horizon tool, it fixes the distortion (mostly) and rotates the image back 90 degrees.
Is there something I am missing here in my process?
On a different, but related note, the forum image upload only supports vertical panos to 3000 px, whereas horizontal panos get 4000 px. That is a bit biased!
Cheers
Tim
Offline
Although I don't do vertical panoramas that often, I have found a similar problem of unwanted rotation back to the horizontal: this is when I do an optimisation in the Control Points Editor.
This of course means I have potentially one more work step, to rotate the image back to the vertical - for every optimisation I do...
btw: rotation of the panorama in the Panorama Editor does not result in rotation of the source images displayed in the Control Points Editor. I wonder, for comment, whether there are occasions when a matching rotation would make working with control points, in the CPE, easier.
Last edited by leedsjoe (2009-02-21 07:03:03)
Offline
You have to render it sideways and then rotate in an image editor afterwards. APP seems to rotate the source images but not the projection... I wonder if it wouldn't be better if APP rotated the projection instead, so we could make vertical panos without having to look sideways...
Things like finetuning rotation and setting verticals have to be done last, since as LeedsJoe said, using "optimize" in the CP Editor destroys our changes. I opened a feature request a long time ago for the "optimise" button to also remember the pano yaw, pitch and roll, so that if, for example, APP automatically detects a 360°x180° pano but shows it tilted at 45° roll, then we can straighten it out and even after using optimise it will still be straightened out and not jump back to 45° roll.
Offline
DrSlony wrote:
I opened a feature request a long time ago for the "optimise" button to also remember the pano yaw, pitch and roll, so that if, for example, APP automatically detects a 360°x180° pano but shows it tilted at 45° roll, then we can straighten it out and even after using optimise it will still be straightened out and not jump back to 45° roll.
You have my vote.
Online
leedsjoe wrote:
btw: rotation of the panorama in the Panorama Editor does not result in rotation of the source images displayed in the Control Points Editor. I wonder, for comment, whether there are occasions when a matching rotation would make working with control points, in the CPE, easier.
Yes, that would be very useful. I'd have thought it would be easy to implement a user control for rotation in the CPE?
Another feature I would like in the CPE would be to boost gamma (or equiavalent) on a temporary basis to make it easier to spot matching areas in underexposed sections of images (IIRC PTGui has that?).
Last edited by mediavets (2009-02-21 15:05:33)
Online
Hmm I never needed that since we cant manually place CPs in APP (yet), but if we do get the ability to do that in APP2 then you got my vote for a gamma slider :]
ad. my previous post: yaw, pitch, roll & center point should be remembered, i forgot to write center point... although perhaps the center point is a result of yaw, pitch and roll... not sure.
Offline
Beside linear projection, any other curent APP projection mode is not "isotrop" (http://en.wikipedia.org/wiki/Isotropy.) Spherical and cylindrical projections don't handle vertical dimension and horizontal dimension in the same way (while linear projection does.)
This is probably preferable because it corresponds to the way we usualy look at things (just try to cross a city why looking toward the zenith !) Stitching very tall panoramas usualy works better with cylindrical or spherical projections when the central vertical line take the place of the horizon. The lines which are vertical in the subject are slightly curved in the process but when the pano is narrow enough in its correct orientation this is hardly visible. The distances between lines which are horizontal in the subject are preserved, that is the upper most stories of a skyscraper are reduced in size according to their distance from the camera (the apparent height of the top part of the tower is preserved in the 90° rotated cylindrical image, the 4rth one from left to right in the example.)
This subject is touched up on in the Cylindrical Horizontal paragraph of Understanding Projecton Modes on this wiki page: http://www.autopano.net/wiki/action/vie … ting_Modes![]()
In a few words: the main problem don't derive from Autopano but from the way we are used to look at things around us. Trying both orientations and deciding the one you like the best is the way to go...
Last edited by GURL (2009-02-21 15:09:20)
Offline
DrSlony wrote:
Hmm I never needed that since we cant manually place CPs in APP (yet), but if we do get the ability to do that in APP2 then you got my vote for a gamma slider :]
Might it not make it easier for SIFT to find CPs too?
Online
Hi All
I have rendered the panoramas and turned them in Ps. That much is easy to do. But the real problem is that with the panorama being sideways you cannot use the vertical lines tool. Also the image is processed differently with the horizontal lines of the final image being straight and the verticals being curved.
I feel this is a serious failing of the software. Now I realise that fewer users use the software for vertical panos, but perhaps more would if this were fixed. The other thing that APP does sometimes with vertical panoramas, that I have not illustrated, is splitting the panorama down the middle at the top and spreading it out in a sort of mushroom shape to fit some (presumably) predefined horizontal proportion. Do others get that as well?
These are not criticisms of the developers, but a genuine desire to improve either my techniques, or the software. Hope you can help.
Cheers
Tim
Offline
mediavets wrote:
DrSlony wrote:
Hmm I never needed that since we cant manually place CPs in APP (yet), but if we do get the ability to do that in APP2 then you got my vote for a gamma slider :]
Might it not make it easier for SIFT to find CPs too?
From a computer's point of view as long as there is a numerical difference between pixels then they're different, no matter whether that difference is visible with human eyes or not. Boosting gamma changes only whats already there, which might make it easier for us to see differences in dark areas of an image, but the differences were already there in the first place.
ps. of course SIFT could be tuned to pay more attention to areas in the middle of the histogram so in that case adjusting gamma would influence detectability, but if that is the case then there is a good reason why it is tuned that way, and if we cheat during detection by changing gamma from e.g. 1.0 to 1.8 and force APP to pay more attention to dark areas (dark areas not discernible to human sight at gamma=1.0 but discernible at gamma=1.8), then light areas (discernible to human sight both at gamma=1.0 and 1.8) might not get the attention they need and the result wont be any better when after detection we change back from 1.8 to 1.0 :] Which is why I wrote that I see it as useful only if we can manually place CPs at specific points.
Offline
Tim.Lewis: but you can use the verticals tool, at least I could in the few vertical panos I did. You just have to first set verticals properly and have a 90° rotation as the last things you do. I agree it would be better if we didn't have to jump these hoops, but doing what you want is entirely possible as it is. Another similar hoop is the one I mentioned earlier about "optimize" destroying our pano alignment. They are both avoidable if you keep to a step by step workflow - first just roughly straighten out the pano after it is detected, then fix all CPs that need fixing, optimize, and then finetune verticals as the last thing you do before setting color anchors and rendering.
If these two things were fixed then we'd be completely free in our workflow order.
Offline
What do you fine tune the verticals with when they are horizontal?
I haven't used colour anchors.
Cheers
Tim
Offline
Powered by PunBB
© Copyright 2002–2005 Rickard Andersson
|
CHOOSING KOLOR Why choose Kolor? Which solution to choose? Download a trial Where can I buy? Education |
SOFTWARE Autopano Pro Autopano Giga Panotour Panotour Pro XnView |
ACCESSORIES Training DVD Panobook PROJECTS Paris 26 Gigapixels Yosemite 17 Gigapixels |
COMMUNITY Forums YouTube channel Google+ |
COMPANY Blog About Kolor Resellers Contact Visit us |
PRESS Press center Press review TOOLS My account |
