cylinder (not cylindrical) projection

[Battle]

I would like to shoot a high resolution panorama of a circular room with vertical walls in rows moving from the floor to the ceiling.  The ideal way to do this is to place the camera in the center and pan around the room for a complete row of images the full 360 degrees around, and then move the camera vertically for the next row of images.  This is effectively a partial matrix panorama.  

In my test photos because distortion factors are uniform in both x and y if I use a rectilinear projection I get many small stitch errors where the photos just won't map correctly because they are being distorted uniformly on both the x and y axes.  A cylindrical projections won't work because its mapping a sphere onto the cylinder.  I need in this example to be able to distort the image along the x axis to straighten out the lines, but not along the y axis.  Although the y axis might have to be shortened or lengthened without moving points on the vertical line along the x axis.   

Any thoughts on whether hugin could be made to do this, and how complicated (level of effort) it would be to add another projection, and if the math to do this pre-exists? Or if another platform has a cylinder projection.

Thanks,

Battle

[bugbear]

Battle wrote:
> I would like to shoot a high resolution panorama of a circular room with vertical walls in rows moving from the floor to the ceiling. The ideal way to do this is to place the camera in the center and pan around the room for a complete row of images the full 360 degrees around, and then move the camera vertically for the next row of images. This is effectively a partial matrix panorama.

Unless the room is tremendously high (like a ruined, empty, castle tower) I don't the
advantage of this over a normal "single point" spherical high res pano. A single point (NPP)
pano can still be extremely high res, if you use a long lens.

BugBear

[Bruno Postle]

On 24 August 2015 at 22:55, Battle wrote:
> I would like to shoot a high resolution panorama of a circular room with
> vertical walls in rows moving from the floor to the ceiling. The ideal way
> to do this is to place the camera in the center and pan around the room for
> a complete row of images the full 360 degrees around, and then move the
> camera vertically for the next row of images. This is effectively a partial
> matrix panorama.

I would stitch each of the rows as a separate cylindrical panorama,
then create a new project to stitch these into a single cylindrical
result - You can optimise the 'e' lens parameter separately for each
of the input cylindricals to align them vertically.

--
Bruno

[Battle]

yes, the room is quite tall.  No way to see more than about 20% from a single point.  So shooting a matrix type is necessary.  

[Battle]

Bruno, Thanks for this tip.  I'll give this a try and report back.  Perhaps I should have been more specific in my original post.  The objective is to unfold/unwrap the cylinder to a flat "rectilinear" style view.  The issue I was having with test images is small blend errors were occurring at the edge of every image where it overlapped with the next image due to the parallax error induced by a matrix shoot.  Hopefully centering a single row in the cylindrical projection will resolve this so that when projects are merged everything will match up.  

I will report back.  Thanks.  

--

Battle

[Battle]

Bruno,

The approach of using multiple projects for a cylindrical projection along the equator, and then combining projects using d & e lens center shift is not working.  There are still blending errors out from the equator.  I think the curved surface of the room with the height of the rows (landscape view) is creating a curved distortion where straight circumference lines at the outer edges (top and bottom) turn in toward the equator the further they are (left and right) from the center of the image.  

Unless I'm missing something this is an unusual distortion pattern that hugin cannot correct for without a special projection for dealing with curved surfaces running in only one direction (cylinder vs sphere) in the original image.  

I would appreciate further thoughts on how to deal with this.     

Regards,

Battle

On Tuesday, August 25, 2015 at 7:07:47 AM UTC-4, Bruno Postle wrote:

[Bruno Postle]

On Wed 02-Sep-2015 at 08:53 -0700, Battle wrote:
>The approach of using multiple projects for a cylindrical projection along
>the equator, and then combining projects using d & e lens center shift is
>not working. There are still blending errors out from the equator. I
>think the curved surface of the room with the height of the rows (landscape
>view) is creating a curved distortion where straight circumference lines at
>the outer edges (top and bottom) turn in toward the equator the further
>they are (left and right) from the center of the image.

It's difficult to say without seeing an example of the problem. A
panorama shot from the centre of a cylindrical building and rendered
in cylindrical projection should be an exact unrolling of the wall
surface.

If the photos were not taken from the centre of the cylinder then
there would be problems (they would actually be in a pannini
projection, but we don't have this as an input option).

--
Bruno

[Battle]

Unfortunately due to architectural constraints the camera is not precisely in the center of the cylinder.  Dealing with a half cylinder for the moment the camera was out of center about 10-12% of diameter in one direction, and less than +-5% 90 degrees the other direction.  And to introduce more variation there is some variance between rows.  Each row assembles seamlessly.  But there are all kinds of errors that appear when multiple rows are combined. Based on your comment about pannini not being available as an input it would seem that I am at a dead end as far as building a single composite image.  If you feel seeing an example would still be useful, I can provide one, but I don't want to post that publicly.  I'd need to know what you'd want to see.  

Thanks,

[Carlos Eduardo G. Carvalho (Cartola)]

Wouldn't a mosaic mode work on it?

Carlos E G Carvalho (Cartola)
http://cartola.org/360
http://www.panoforum.com.br/

--
A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ
---
You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hugin-ptx+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hugin-ptx/918639c3-e5a2-4c50-aaaf-f17651826bfe%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Bruno Postle]

On 3 September 2015 17:30:40 BST, Battle wrote:
>Unfortunately due to architectural constraints the camera is not
>precisely
>in the center of the cylinder. Dealing with a half cylinder for the
>moment
>the camera was out of center about 10-12% of diameter in one direction,
>and
>less than +-5% 90 degrees the other direction. And to introduce more
>variation there is some variance between rows. Each row assembles
>seamlessly. But there are all kinds of errors that appear when
>multiple
>rows are combined.

I think you ought to be able to get a result with this amount of variation so long as you have enough photos.

You can send me an output image or a project file and I'll see if there is anything obvious that I can suggest.

--
Bruno

[Battle]

Bruno,

I've sent you an email directly with a link to download files.  

Thanks,

Battle

[Battle]

Carlos, a mosaic was the first thing I tried.  I have a fair amount of experience with mosaics, but mostly on planar surfaces using rectilinear projection.  One things I've noticed on projects like that in the past, is that sometimes the variables get so locked into a configuration that just won't stitch that I have to reset and start over.  Often a matrix pano I'm struggling with will stitch correctly if I can get the XY parameters close, with everything else reset to zero and then optimize.  I may go back and try this for a particularly flat area that I think hugin should be able to unwrap.