How to best shoot and stitch this?

[Terry Duell]

Hello All,
The attachment shows the Vietnam Veterans Commemorative Walk, in Seymour,
Victoria.
It is approx. 80-ish metres long, with about 52 glass panels on each side,
each panel approx 2m high, 1.5m wide.
The panels are etched with images from the war, and the names of all the
Australian veterans.
Shooting and stitching a pano of each side looks like a tricky project.
The right side of the attached image is roughly north, so shooting the
south side is always going to present a bit of a problem with the light,
i.e. the sun is always to the north.
On the north side the panels have much better light, but there is
vegetation about 4 to 5m from the panels.
Standing on the edge of the vegetation, a 24mm lens (36mm equiv) shooting
approx normal to the panels gives about 2.5 panel wide coverage.
Does anyone have any comments on whether a pano might be possible, and any
ideas on how to tackle it?

Cheers,
--
Regards,
Terry Duell

[Roger Goodman]

Terry,
Have you considered using a flash (or two) to lighten the south
side panels? With a bright flash about 2 meters from the camera, you
should get some good lighting. Just a thought.
Roger

[Terry Duell]

Hello Roger,

On Tue, 04 Mar 2014 12:59:08 +1100, Roger Goodman <rlgo...@cox.net>
wrote:

> Terry,
> Have you considered using a flash (or two) to lighten the south
> side panels? With a bright flash about 2 meters from the camera, you
> should get some good lighting. Just a thought.

That's not really the main issue. Perhaps I didn't explain my concerns
well enough.
For a normal single camera position pano, it isn't possible to find one
location that will capture the whole length, and provide a good view of
each the panels on one side of the walk.
Because of the considerable curve of the walk I don't think it is possible
to shoot it as a linear pano...i.e. many different camera positions. There
is a continual change in Trz value, which is small from one panel to the
next and may allow a linear pano stitch of 2 images, each with 2 panels,
overlapping one panel for a 3 panel linear stitch, but would I then be
able to stitch all these linear panos together?
One other approach I have pondered on, is to shoot a couple of panels to
my right and and a couple to my left and stitch as normal, and make about
12 or 13 of these sub-panos, each with one panel overlap on the ends with
the next sub-pano. Then try to stitch these normal sub-panos together
using the one panel overlap and do these as linear panos, but I'm not sure
that the overlap will be enough.
So, not really sure if any of the above has any hope of success.

Anyone attempted anything similar?

[Terry Duell]

On Tue, 04 Mar 2014 13:25:35 +1100, Terry Duell <tdu...@iinet.net.au>
wrote:

[snip]

> Because of the considerable curve of the walk I don't think it is
> possible to shoot it as a linear pano...i.e. many different camera
> positions. There is a continual change in Trz value, which is small from
> one panel to the next and may allow a linear pano stitch of 2 images,
> each with 2 panels, overlapping one panel for a 3 panel linear stitch,
> but would I then be able to stitch all these linear panos together?
> One other approach I have pondered on, is to shoot a couple of panels to
> my right and and a couple to my left and stitch as normal, and make
> about 12 or 13 of these sub-panos, each with one panel overlap on the
> ends with the next sub-pano. Then try to stitch these normal sub-panos
> together using the one panel overlap and do these as linear panos, but
> I'm not sure that the overlap will be enough.
> So, not really sure if any of the above has any hope of success.

The more I think about this, the more I am convincing myself that none of
the above can produce a reasonable result.

[Marius Loots]

Hallo Terry,

Tuesday, March 4, 2014, 4:25:35 AM, you wrote:
Terry> the next sub-pano. Then try to stitch these normal sub-panos together
Terry> using the one panel overlap and do these as linear panos, but I'm not sure
Terry> that the overlap will be enough.
Terry> So, not really sure if any of the above has any hope of success.
Terry> Anyone attempted anything similar?

In 2006 I took a set of similar photos of the Korean War Veterans
Memorial in Washington, DC. With the intent to stitch the panels.

I unfortunately did not have a lot of time to spend on photographing
the sections, so did a very quick "scan" of the 50 m wall. It was
freehand, staying at approximately the same distance from the wall and
stepping sideways roughly the same distance between shots for a total
of 65 shots.

At that time I attempted to stitch the images, using the same
technique as that used for scans. I can't remember how far I got, but
eventually gave up, mostly because of the limits of my pc at the time.

My biggest problem was control points. The granite is highly
reflective, which meant that I showed up in each photograph and
control points were added automatically linking up myself in all the
images. The same with all the images engraved in the granite. I have
now retrieved the photos and will gave it another attemp today. Let
you know how it went.

Groetnis
Marius
mailto:mlo...@medic.up.ac.za
--
add some chaos to your life and put the world in order
http://www.mapungubwe.co.za/
http://www.chaos.co.za/
skype: marius_loots

Hierdie boodskap en aanhangsels is aan 'n vrywaringsklousule
onderhewig. Volledige besonderhede is by
www.it.up.ac.za/documentation/governance/disclaimer/
beskikbaar.

[Brandan]

Hi there,

I have never tried one of what you are describing so I have no insights into the stitching. For the actually taking the photos can you shoot the photos on a cloudy day, or maybe yearly or late in the day when you have indirect lighting? I am hopeful that if you could get the right light it would cut down on the reflection that Marius Loots has in his photos. Which would have to help with the control point finding.

[Terry Duell]

Hello Brandan,

On Tue, 04 Mar 2014 19:08:37 +1100, Brandan <bra...@flyingtsalers.com>
wrote:

> I am hopeful that if you could get the right light it
> would cut down on the reflection that Marius Loots has in his photos.
> Which would have to help with the control point finding.
>

I have taken a number of shots of the walk, and auto control point finding
can be a bit problematic, but it isn't too hard to manually add them. I
don't think reflection is major issue.
I think the main issue is that it is probably impossible to stitch a pano
that gives a clear view of all panels.

Thanks for your comments.

[Terry Duell]

Hello Marius,


On Tue, 04 Mar 2014 17:57:03 +1100, Marius Loots <mlo...@medic.up.ac.za>
wrote:

> My biggest problem was control points. The granite is highly
> reflective, which meant that I showed up in each photograph and
> control points were added automatically linking up myself in all the
> images. The same with all the images engraved in the granite. I have
> now retrieved the photos and will gave it another attemp today. Let
> you know how it went.
>

I don't have the same problems with reflections. The really big issue, I
think, is the size and shape.
The size precludes a normal pano, and the shape precludes a linear pano.

This might be a silly idea, but is there any post processing you can do on
the images to counter the reflections that will leave detail that CPFind
can use to get control points? If so, then save your .pto and substitute
your original filenames and restitch.
Please let us know how you get on with your new attempt.

[paul womack]

Terry Duell wrote:
> Hello Brandan,
>
> On Tue, 04 Mar 2014 19:08:37 +1100, Brandan <bra...@flyingtsalers.com> wrote:
>
>> I am hopeful that if you could get the right light it
>> would cut down on the reflection that Marius Loots has in his photos. Which would have to help with the control point finding.
>>
> I have taken a number of shots of the walk, and auto control point finding can be a bit problematic, but it isn't too hard to manually add them.

I don't think I've ever used auto control point creation. When I've tried it
I always spend more time editing points, and deleting bad ones,
than it takes me to create nigh perfect ones from scratch.

BugBear

[paul womack]

Terry Duell wrote:
> Hello All,
> The attachment shows the Vietnam Veterans Commemorative Walk, in Seymour, Victoria.
> It is approx. 80-ish metres long, with about 52 glass panels on each side, each panel approx 2m high, 1.5m wide.

.

.
> Does anyone have any comments on whether a pano might be possible, and any ideas on how to tackle it?

It's clearly possible - if nothing else it would be possible, if tedious,
to simply take 52 shots, each centred on a single panel,
pick a single panel as a reference, and correct this panel
to be perfectly proportioned and rectilinear.

http://hugin.sourceforge.net/tutorials/perspective/en.shtml

Then create 52 2-frame panos, each with the reference
panel as one picture, and a fresh picture as the other,
and correct the shot to the reference panel.

Stitch each panorama with the reference panel invisible, and the
panel precisely cropped.

The resulting images will tessalate perfectly.

BugBear

[kfj]

If you can live without the curve of the wall, I would do this:

- In a drawing program like gimp, create a line drawing of a stripe with sections having the same aspect ratio as the individual images on the wall - imagine a ladder put down horizontal with rungs upright

- Mask the individual normal shots so that there is some of the left and right neighbouring image visible; enough to let the blending software create an invisible blend.

- 'pin' the four corners of each image to the corresponding points on your skeleton drawing (the ladder)

- use the ladder drawing as reference and optimize the other images for r, p, y, x, y and z. If the otimization fails, you could try adding line control points along the ladder's lines; in my experience with stitching mosaics from handheld shots this has always helped getting the optimization to work and achieving a good result.

- leave the placement of the seams to enblend; the curvature might be slight enough not to create too much parallactic error

This process might be less work than processing each individual image to be perfectly normal and of the right size and then mosaicing them together. And making a trial with just a few images won't cost you too much time.

Kay

[Marius Loots]

Hallo Terry,

Tuesday, March 4, 2014, 10:50:49 AM, you wrote:
>> My biggest problem was control points. The granite is highly
>> reflective, which meant that I showed up in each photograph and
>> control points were added automatically linking up myself in all the
>> images. The same with all the images engraved in the granite. I have
>> now retrieved the photos and will gave it another attemp today. Let
>> you know how it went.
>>

Terry> I don't have the same problems with reflections. The really big issue, I
Terry> think, is the size and shape.
Terry> The size precludes a normal pano, and the shape precludes a linear pano.

Terry> This might be a silly idea, but is there any post processing you can do on
Terry> the images to counter the reflections that will leave detail that CPFind
Terry> can use to get control points? If so, then save your .pto and substitute
Terry> your original filenames and restitch.
Terry> Please let us know how you get on with your new attempt.

I have uploaded my next attempt at:
http://www.flickr.com/photos/marius_loots/12927837453/

One correction, it wasn't my reflection giving so much problems, but
the reflections of trees that are repeated across the whole section.
But, the 2012 version of hugin CPfind deals a lot better with this
than whatever I used with the version current in 2006.

The result is still relatively uneven, even with horisontal and
vertical control points added and correcting for barrel distortion.
The barrel distortion being quite visible in the original images. I
think if one has a tripod and keep the distance from the wall constant
using a tape measure, the results would be a lot better. In the case
of the curved wall one would maybe also keep the angle towards the
panel the same.

I will give it another go, using the various suggestions that has been
made. In summary, I think it would be possible, given a little more
time on the photographs than I took (five minutes for 66 photos of a
50 meter stretch).

Will update on the revised attempts.

[dgjoh...@accesscomm.ca]

My best suggestion would be to make sure that you are always at 90 degree to the tangent of the centre point of the section of the wall you're photographing. And have lots of overlap so you only use the vertical centr of each image.

Sent from my BlackBerry® wireless handheld

--
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/op.xb6mckmjrs0ygh%40localhost.
For more options, visit https://groups.google.com/groups/opt_out.

[Terry Duell]

Hello Paul,

On Tue, 04 Mar 2014 20:43:24 +1100, paul womack <pwo...@papermule.co.uk>
wrote:

>
> It's clearly possible - if nothing else it would be possible, if tedious,
> to simply take 52 shots, each centred on a single panel,
> pick a single panel as a reference, and correct this panel
> to be perfectly proportioned and rectilinear.
>
> http://hugin.sourceforge.net/tutorials/perspective/en.shtml
>
> Then create 52 2-frame panos, each with the reference
> panel as one picture, and a fresh picture as the other,
> and correct the shot to the reference panel.
>
> Stitch each panorama with the reference panel invisible, and the
> panel precisely cropped.
>
> The resulting images will tessalate perfectly.
>

Thanks for your thoughts on this.
As you say, possible, but tedious.

[Terry Duell]

Hello Kay,

On Tue, 04 Mar 2014 21:19:46 +1100, kfj <_k...@yahoo.com> wrote:


>
> If you can live without the curve of the wall, I would do this:
>
> - In a drawing program like gimp, create a line drawing of a stripe with
> sections having the same aspect ratio as the individual images on the
> wall - imagine a ladder put down horizontal with rungs upright
>
> - Mask the individual normal shots so that there is some of the left and
> right neighbouring image visible; enough to let the blending software
> create an invisible blend.
>
> - 'pin' the four corners of each image to the corresponding points on
> your skeleton drawing (the ladder)
>
> - use the ladder drawing as reference and optimize the other images for
> r, p, y, x, y and z. If the otimization fails, you could try adding line
> control points along the ladder's lines; in my experience with stitching
> mosaics from handheld shots this has always helped getting the
> optimization to work and achieving a good result.
>
> - leave the placement of the seams to enblend; the curvature might be
> slight enough not to create too much parallactic error
>
> This process might be less work than processing each individual image to
> be perfectly normal and of the right size and then mosaicing them
> together.
> And making a trial with just a few images won't cost you too much time.
>

Thanks for that. I'll have a try and see how it goes.

[Terry Duell]

Hello Kay,

Sorry if I am a bit dense, but a couple of questions on detail, if I may.

On Tue, 04 Mar 2014 21:19:46 +1100, kfj <_k...@yahoo.com> wrote:

[snip]

>
> If you can live without the curve of the wall, I would do this:
>
> - In a drawing program like gimp, create a line drawing of a stripe with
> sections having the same aspect ratio as the individual images on the
> wall
> - imagine a ladder put down horizontal with rungs upright
>
> - Mask the individual normal shots so that there is some of the left and
> right neighbouring image visible; enough to let the blending software
> create an invisible blend.
>
> - 'pin' the four corners of each image to the corresponding points on
> your skeleton drawing (the ladder)

Where is this to be done, in Gimp or Hugin?
Can you be a bit more explicit as to what you mean by 'pin' an image to
another, it isn't a term I see used in Gimp or Hugin.

>
> - use the ladder drawing as reference and optimize the other images for
> r, p, y, x, y and z.

What is the ladder drawing in this context...the bare drawing, or the
drawing with all the images 'pinned'?
By 'reference' do you mean (in hugin terminology) the 'anchor' for
position?

[kfj]

On Wednesday, March 5, 2014 10:00:48 AM UTC+5:30, Tduell wrote:

Hello Kay,

Sorry if I am a bit dense, but a couple of questions on detail, if I may.

On Tue, 04 Mar 2014 21:19:46 +1100, kfj <_k...@yahoo.com> wrote:

> - 'pin' the four corners of each image to the corresponding points on  
> your skeleton drawing (the ladder)

Where is this to be done, in Gimp or Hugin?
Can you be a bit more explicit as to what you mean by 'pin' an image to

another, it isn't a term I see used in Gimp or Hugin.

Okay, sorry for the sloppy terminology. By 'pinning' I mean setting control points between a reference image with reference points and corresponding points in partial images which need placement. Imagine, for example, having a set of image tiles and a reference image showing a grid. Pinning would then mean that you set CPs from the four corners of each tile to the grid points where the image is to be placed. Another use of pinning was my method of fixing prominent features in landscape panoramas to an artificial panorama which is used as reference, resulting in what I labeled a 'canonical' panorama with near-perfect horizon and orientation, see

https://groups.google.com/forum/#!topic/hugin-ptx/l32L3Iffv6w

> - use the ladder drawing as reference and optimize the other images for  
> r, p, y, x, y and z.

What is the ladder drawing in this context...the bare drawing, or the  
drawing with all the images 'pinned'?

The ladder drawing, in this context, is the line drawing made in gimp which serves as the surface you 'pin' your individual images to.

By 'reference' do you mean (in hugin terminology) the 'anchor' for  
position?

Yes, that's what I mean. You are right, anchor is the hugin term, or to be more precise, position anchor. And when you finally stitch the panorama, you simply deactivate this anchor image, resulting in a panorama from the individual photos, held by the invisible 'spine' of the hidden artificial anchor image.

The 'pinning' technique has quite a few interesting applications, but I think it isn't used much, since it doesn't occur to many people to mix in images which don't visibly contribute to the final product. Another example of pinning is using a reference showing an azimuthal grid, possibly with the sun's position marked which can be found by using the techniques discussed here:

https://groups.google.com/forum/#!topic/hugin-ptx/vWRO7SxdFxQ

If your set of photographs show both the nadir and the sun, you are just a few clicks away from a level horizon and a good orientation.

Kay

[Terry Duell]

Hello Kay,

On Wed, 05 Mar 2014 18:55:55 +1100, kfj <_k...@yahoo.com> wrote:

[snip]

>> Can you be a bit more explicit as to what you mean by 'pin' an image to
>> another, it isn't a term I see used in Gimp or Hugin.

>
> Okay, sorry for the sloppy terminology. By 'pinning' I mean setting
> control
> points between a reference image with reference points and corresponding
> points in partial images which need placement. Imagine, for example,
> having
> a set of image tiles and a reference image showing a grid. Pinning would
> then mean that you set CPs from the four corners of each tile to the grid
> points where the image is to be placed.

[snip]

>>
>> What is the ladder drawing in this context...the bare drawing, or the
>> drawing with all the images 'pinned'?
>>
>
> The ladder drawing, in this context, is the line drawing made in gimp
> which serves as the surface you 'pin' your individual images to.
>

>> By 'reference' do you mean (in hugin terminology) the 'anchor' for
>> position?
>
>
> Yes, that's what I mean. You are right, anchor is the hugin term, or to
> be more precise, position anchor. And when you finally stitch the
> panorama,
> you simply deactivate this anchor image, resulting in a panorama from the
> individual photos, held by the invisible 'spine' of the hidden artificial
> anchor image.
>

Thanks for the clarification.

[Emad ud din Bhatt]

for reflection problen u can use a circular polarizer filter on lens. rotate filter to cut reflections.
rest of discussion is very difficult to digest. well i would try to shoot with excessive overlap say 70-80% and use only 30-20% of central strip to stitch.

--
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+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hugin-ptx/op.xb8ru0avrs0ygh%40localhost.

[Terry Duell]

Hello Emad,

On Mon, 10 Mar 2014 15:04:10 +1100, Emad ud din Bhatt <xyz...@gmail.com>
wrote:

> for reflection problen u can use a circular polarizer filter on lens.
> rotate filter to cut reflections.
> rest of discussion is very difficult to digest. well i would try to shoot
> with excessive overlap say 70-80% and use only 30-20% of central strip to
> stitch.

Yes, a polarising filter is an option for reflections.
I really haven't thought a lot about that aspect thus far, the non planar
surface and inability to get it all from one location being the big
hurdles.
I think Kay's approach seems to have some promise, and I hope to be able
to try it soon.

Thanks for your help.