Alignment tutorial reworked

[Greg 'groggy' Lehey]

I've just brought the tutorial on aligning multiple photos
(http://wiki.panotools.org/Align_a_stack_of_photos) up to date.
Review welcome.

I've also written a much more detailed version at
http://www.lemis.com/grog/photography/aligning-with-Hugin.php
If anybody feels like incorporating any of that into the tutorial,
feel free to do so. I just found it too complicated to fit into the
Wiki mould.

Greg
--
Sent from my desktop computer.
Finger gr...@FreeBSD.org for PGP public key.
See complete headers for address and phone numbers.
This message is digitally signed. If your Microsoft MUA reports
problems, please read http://tinyurl.com/broken-mua

[Margaret Wong]

Thanks for this tutorial :) It will be useful as I have ended up doing
such images by hand in gimp!

> I've just brought the tutorial on aligning multiple photos
> (http://wiki.panotools.org/Align_a_stack_of_photos) up to date.
> Review welcome.
>
> I've also written a much more detailed version at
> http://www.lemis.com/grog/photography/aligning-with-Hugin.php
> If anybody feels like incorporating any of that into the tutorial,
> feel free to do so. I just found it too complicated to fit into the
> Wiki mould.
>
> Greg




--

kind regards

Margaret

[Greg 'groggy' Lehey]

On Sunday, 10 February 2013 at 14:47:14 +1100, Greg 'groggy' Lehey wrote:
> I've just brought the tutorial on aligning multiple photos
> (http://wiki.panotools.org/Align_a_stack_of_photos) up to date.
> Review welcome.

Well, "Thomas" responded in a rather unusual way, by changing the
documentation further without discussion. Unfortunately, the changes
mean that the instructions no longer work. I don't know who he is; he
doesn't have a user page, and there are multiple Thomases on the list.
Thomas, you could at least have discussed the changes before making
them.

In detail:

- The instructions now refer to the new GUI, which is still not in a
released version of Hugin. For normal users this is confusing.

- They recommend align_image_stack instead of panomatic. When I tried
this on the example images, it didn't find any control points
between the first and third image. It's also much slower than
panomatic.

- He changed the selection of optimization from "Positions and View
(y,p,r,v)" to "Positions (y,p,r)". It's not clear what advantage
this should bring, but this doesn't work with images of differing
sizes, like the example images I describe. In my test, it gave me
an average control point distance of 5.46 and a maximum of 12.28.
"Positions and View (y,p,r,v)" gave me an average of 0.31 and a
maximum of 1.03.

On the positive side, changing the cropping to HDR autocrop (now
incorporated into my tutorial at
http://www.lemis.com/grog/photography/aligning-with-Hugin.php)
improves things greatly. But it's the only good change.

This all goes to show that it's a good idea to test things before
committing them. Discussion is also a good thing. Thomas, would you
please identify yourself and comment?

[Harry van der Wolf]

Hi Greg,

2013/2/11 Greg 'groggy' Lehey <groo...@gmail.com>

Thomas is Thomas Modes.

 

I again changed the wiki page to reflect both the new gui and the old gui.

 

Please check and change accordingly.

 

Harry

[T. Modes]

Hi Greg,

> In detail:
>
> - The instructions now refer to the new GUI, which is still not in a
>   released version of Hugin.  For normal users this is confusing.
>

The wiki page are also used as help files and shipped with Hugin. If
the new version would be shipped without the updated pages, it would
also confusing. And we can't update all pages at once. So we started
with the update of the wiki pages. The first beta release will follow
in the next weeks. So there is a short time, where wiki and release
are out of date.
But every hugin release comes with its own help, which contains the
information matching the installed version. So we take this short time
of not sync wiki and release as acceptable.

> - They recommend align_image_stack instead of panomatic.  When I tried
>   this on the example images, it didn't find any control points
>   between the first and third image.  It's also much slower than
>   panomatic.

Many control point detector work for this case. Depending on the
images one is better than the other. So I added a link for other
detectors instead of the your preferences for one specific cpg.
Align_image_stack does always link between consecutive images. So it
does not match image 1 with 3. This is a feature (necessary e.g. for
focus stacks) and not a bug ;-)

>
> - He changed the selection of optimization from "Positions and View
>   (y,p,r,v)" to "Positions (y,p,r)".  It's not clear what advantage
>   this should bring, but this doesn't work with images of differing
>   sizes, like the example images I describe.  In my test, it gave me
>   an average control point distance of 5.46 and a maximum of 12.28.
>   "Positions and View (y,p,r,v)" gave me an average of 0.31 and a
>   maximum of 1.03.
>

If you have only one image / image stack the fov can't be correctly
fitted. There are not enough information to do it. For different
(fixed) fov you can always get the (nearly) same errors.
This may work, but it may also create havoc.
Normally all images of the stack are shoot with the same fov. So fov
should not be fitted. In your special case - which you did not
mentioned in the wiki - you can fit the fov of the second lens. But it
may be a good idea to leave the fov of image 0 fixed. The setting
"Positions and view" will always fit fov of all images. This is not
well suited for this case.

Also align_image_stack (when using direct with images) leaves the fov
of the first image fixed.

Thomas

[Greg 'groggy' Lehey]

On Monday, 11 February 2013 at 11:02:55 -0800, T. Modes wrote:
> Hi Greg,
>
>> In detail:
>>
>> - The instructions now refer to the new GUI, which is still not in a
>>   released version of Hugin.  For normal users this is confusing.
>
> The wiki page are also used as help files and shipped with Hugin. If
> the new version would be shipped without the updated pages, it would
> also confusing.

Agreed. When the new version is released, it will need updated
documentation.

> And we can't update all pages at once.

No, but that doesn't mean you need to publish the new pages before
they make sense.

> So we started with the update of the wiki pages. The first beta
> release will follow in the next weeks. So there is a short time,
> where wiki and release are out of date.

They'll be out of date for ever for people who don't want to update to
the new GUI.

> But every hugin release comes with its own help, which contains the
> information matching the installed version.

The current version didn't come with this information. It was years
out of date.

> So we take this short time of not sync wiki and release as
> acceptable.

Do the users?

The obvious thing to do here is to update the page to describe both
scenarios, and also to discuss some of the tradeoffs of which I wasn't
aware (read on). I'll do this unless there's some objection.

>> - They recommend align_image_stack instead of panomatic.  When I tried
>>   this on the example images, it didn't find any control points
>>   between the first and third image.  It's also much slower than
>>   panomatic.
>
> Many control point detector work for this case. Depending on the
> images one is better than the other. So I added a link for other
> detectors instead of the your preferences for one specific cpg.
> Align_image_stack does always link between consecutive images. So it
> does not match image 1 with 3. This is a feature (necessary e.g. for
> focus stacks) and not a bug ;-)

I've been doing more playing around, and there are many areas here
that are even less straightforward than I thought. For example,
align_image_stack wants all files to be the same size. Otherwise it
dies with an error message that it immediately erases.

> If you have only one image / image stack the fov can't be correctly
> fitted. There are not enough information to do it. For different
> (fixed) fov you can always get the (nearly) same errors. This may
> work, but it may also create havoc. Normally all images of the
> stack are shoot with the same fov. So fov should not be fitted. In
> your special case - which you did not mentioned in the wiki - you
> can fit the fov of the second lens. But it may be a good idea to
> leave the fov of image 0 fixed. The setting "Positions and view"
> will always fit fov of all images. This is not well suited for this
> case.

Yes, I've since discovered this today. It seems to be another "it
depends" case, and I'm still trying to get my head around how to
adjust slightly different images without the v parameter.

[Rogier Wolff]

On Mon, Feb 11, 2013 at 11:02:55AM -0800, T. Modes wrote:
> If you have only one image / image stack the fov can't be correctly
> fitted. There are not enough information to do it. For different
> (fixed) fov you can always get the (nearly) same errors.
> This may work, but it may also create havoc.

> Normally all images of the stack are shoot with the same fov. So fov
> should not be fitted. In your special case - which you did not

In all cameras that I have adjusting the focus when in macro-mode
clearly changes the "magnification" as well. So in my experience,
optimizing "v" makes sense....

> mentioned in the wiki - you can fit the fov of the second lens. But it
> may be a good idea to leave the fov of image 0 fixed.

... but as you say you should leave exactly one fov fixed.

Roger.

--
** R.E....@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2600998 **
** Delftechpark 26 2628 XH Delft, The Netherlands. KVK: 27239233 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
The plan was simple, like my brother-in-law Phil. But unlike
Phil, this plan just might work.