New hugin tutorial - Creating 360° enfused panoramas

364 views
Skip to first unread message

Bruno Postle

unread,
Jan 30, 2008, 7:09:38 PM1/30/08
to Hugin ptx
I've uploaded a new tutorial, though this one is more of a
'walkthrough':

http://hugin.sourceforge.net/tutorials/enfuse-360/

This covers shooting a bracketed 360° scene and creating an exposure
blended (enfused) panorama - All entirely using features in the
upcoming hugin 0.7.0 release.

Any clarifications or suggestions for improvement welcome.

--
Bruno

akse...@gmail.com

unread,
Jan 30, 2008, 11:56:23 PM1/30/08
to hugi...@googlegroups.com
Very good Bruno. That foot must be what was left after someone was
run over by the truck driving on the sidewalk.

I do not understand how the process knew to build each of the pre-
enfused panos from the correct bracket images. That is a question I
have been wondering about and I hoped the tutorial would explain. Was
there a step involving the Photometric section of Camera and Lens
where the EV is entered and then the Link control checked?

I assume one should not run the Exposure Photometric "Optimisation"
process.

Allan

Bruno Postle

unread,
Jan 31, 2008, 10:19:10 AM1/31/08
to Hugin ptx
On Wed 30-Jan-2008 at 22:56 -0600, akse...@gmail.com wrote:
>
>Very good Bruno. That foot must be what was left after someone was
>run over by the truck driving on the sidewalk.

Yes there are quite a few ghosts and extra limbs in that panorama, I
just left the output exactly as it came out without any retouching.

>I do not understand how the process knew to build each of the pre-
>enfused panos from the correct bracket images. That is a question I
>have been wondering about and I hoped the tutorial would explain. Was
>there a step involving the Photometric section of Camera and Lens
>where the EV is entered and then the Link control checked?

hugin uses a 'heuristic' based on the relative positions and
exposure values of the photos, so it works if you have EXIF data or
if you do photometric optimisation.

>I assume one should not run the Exposure Photometric "Optimisation"
>process.

The Align... button in the Assistant does photometric optimisation
as the last step, so I did in this tutorial. Though the EV values
themselves are discarded when rendering the input passed to enfuse.

(but vignetting isn't discarded, which may or may not be an
oversight, someone needs to do some tests)

--
Bruno

Carl von Einem

unread,
Jan 31, 2008, 12:50:45 PM1/31/08
to hugi...@googlegroups.com
Hi Bruno,

thanks for describing your workflow, it's very interesting even if I
don't (or not yet) use enfuse. Your tutorials are great! :-)

Carl

Bob Bright

unread,
Feb 2, 2008, 2:59:23 AM2/2/08
to hugi...@googlegroups.com
Hi Bruno,

Very nice.  I really like the "walkthrough" style -- not only do users get a helpful introduction to enfuse, but they get lots of other useful information about workflow, etc. by looking over your shoulder while you're stitching a real live example.

I think there's one small error, though.  The tutorial states that when you ask for a "Blended panorama (enfuse)" on the hugin Stitcher tab, nona first remaps the images, then enblend creates three complete panos (1 for each exposure setting), and finally enfuse merges the three into a single panorama.  Aren't steps 2 and 3 reversed?  I.e., doesn't hugin fuse each exposure stack before blending the remapped+fused images into a complete pano?

At any rate, this seems to be the way hugin works when I select "Blended panorama (enfuse)" alone prior to stitching.  I assume the behaviour doesn't change when one selects "Blended exposure layers" in addition to "Blended panorama (enfuse)".  (I haven't been able to confirm this since I'm getting crashes when I select both options -- SVN2744 on Ubuntu 7.10.  But if the behaviour does change, I guess it would have to be considered a bug, since fusing prior to blending is the "correct" behaviour.)

Best regards,
BBB

Pablo d'Angelo

unread,
Feb 2, 2008, 3:11:55 AM2/2/08
to hugi...@googlegroups.com
Bob Bright wrote:
> Hi Bruno,

>
> I think there's one small error, though. The tutorial states that when
> you ask for a "Blended panorama (enfuse)" on the hugin Stitcher tab,
> nona first remaps the images, then enblend creates three complete panos
> (1 for each exposure setting), and finally enfuse merges the three into
> a single panorama. Aren't steps 2 and 3 reversed? I.e., doesn't hugin
> fuse each exposure stack before blending the remapped+fused images into
> a complete pano?

Yes, they are reversed. first enfuse stacks then enfuse blended stacks.
This allows one to use a different number of images for each stack.
Otherwise enfuse would leave ugly seams at places where the extra shots are
placed. The drawback is that enfuse can use less layers for exposure
blending and the results might be a little worse.

ciao
Pablo

Bob Bright

unread,
Feb 2, 2008, 5:14:30 AM2/2/08
to hugi...@googlegroups.com
On Sat, 2008-02-02 at 09:11 +0100, Pablo d'Angelo wrote:

Yes, they are reversed. first enfuse stacks then enfuse blended stacks.
This allows one to use a different number of images for each stack. 
Otherwise enfuse would leave ugly seams at places where the extra shots are 
placed. The drawback is that enfuse can use less layers for exposure 
blending and the results might be a little worse.

ciao
   Pablo

Hi Pablo,

I guess this is a theoretical drawback all right, but in practice I really haven't seen much difference in my (admittedly haphazard) testing.

The big practical benefit of enfusing before blending (I would say it's a *huge* benefit, for those of us who were doing manual exposure blending in the gimp or whatever prior to enfuse) is that enblend only gets to run once on a set of images which have already been exposure blended.  So you don't get variations in blending seam placement between the different exposures, which used to lead to frequent ghosting problems in the final exposure-blended pano.

BTW, can you think of any reason not to enfuse the exposure stacks *prior* to processing in hugin?  I realize that doing this destroys hugin's ability to compensate for vignetting, but in my testing so far, whatever vignetting remains in the enfused inputs is pretty easily handled by enblend.  Sometimes I seem to get slightly better final results when I feed pre-enfused images to hugin, and sometimes I get slightly better final results with un-enfused inputs -- but in either case, there's not much to choose between the two.

BBB


Pablo d'Angelo

unread,
Feb 2, 2008, 8:22:07 AM2/2/08
to hugi...@googlegroups.com
Hi all,

Bob Bright wrote:
> On Sat, 2008-02-02 at 09:11 +0100, Pablo d'Angelo wrote:
>
>> Yes, they are reversed. first enfuse stacks then enfuse blended stacks.

Seems that I was still asleep while writing this...
Sorry, replace the enfuse blended stacks with enblend enfused stacks.

The process for the enfused output in hugin is:

1. remap images, do not change exposure but correct vignetting
2. enfuse stacks
3. blend fused stacks with enblend

>> This allows one to use a different number of images for each stack.
>> Otherwise enfuse would leave ugly seams at places where the extra shots are
>> placed. The drawback is that enfuse can use less layers for exposure
>> blending and the results might be a little worse.
>>
>

> The big practical benefit of enfusing before blending (I would say it's
> a *huge* benefit, for those of us who were doing manual exposure
> blending in the gimp or whatever prior to enfuse) is that enblend only
> gets to run once on a set of images which have already been exposure
> blended. So you don't get variations in blending seam placement between
> the different exposures, which used to lead to frequent ghosting
> problems in the final exposure-blended pano.

Yes, this is indeed one of the major advanages. On the other hand it is
possible (with some advanced scripting, or by modifying enblend) to force
the same seams for all panoramas.

The major advantage of enfusing after remapping is that slight
missalignments between the images in a stack can be handled.

> BTW, can you think of any reason not to enfuse the exposure stacks
> *prior* to processing in hugin? I realize that doing this destroys
> hugin's ability to compensate for vignetting, but in my testing so far,
> whatever vignetting remains in the enfused inputs is pretty easily
> handled by enblend.

The major advantage of fusing before remapping is that there are no problems
with the zenith in spherical panoramas. If an extra alignment step is
required, the images will be interpolated one more time, but since the
interpolation for the alignment will not include large scale changes or
distortions, it should not be too destructive.

Note that I haven't tried any of this, its just theoretical thinking, which
should be verified in practice, too.

ciao
Pablo

Ken Warner

unread,
Feb 2, 2008, 9:36:16 AM2/2/08
to hugi...@googlegroups.com

How do you do exposure blending in GIMP?

Seb Perez-D

unread,
Feb 2, 2008, 11:58:33 AM2/2/08
to hugi...@googlegroups.com
On Sat, Feb 2, 2008 at 3:36 PM, Ken Warner <kwa...@uneedspeed.net> wrote:
> How do you do exposure blending in GIMP?

You ask Google

http://www.google.com/search?q=exposure+blending+in+GIMP

Cheers!

Seb

Erik Krause

unread,
Feb 2, 2008, 12:07:24 PM2/2/08
to hugin-ptx
On Saturday, February 02, 2008 at 6:36, Ken Warner wrote:

> How do you do exposure blending in GIMP?

-> http://turtle.as.arizona.edu/jdsmith/exposure_blend.php

best regards
Erik Krause
http://www.erik-krause.de

Ken Warner

unread,
Feb 2, 2008, 12:33:17 PM2/2/08
to hugi...@googlegroups.com
Coool! Thanks Seb and Erik... 'nuther toy to play with...

Bob Bright

unread,
Feb 2, 2008, 2:55:11 PM2/2/08
to hugi...@googlegroups.com
Hi Ken,

I guess it's always good to learn new techniques, but I wouldn't spend too much time playing with exposure blending in the gimp if I were you.  I think you'll find that enfuse delivers superior results with less effort.

(On the other hand, using Smith's exposure-blend plugin might be a good way to fine tune the results of enfusing, if there was something about the latter that you were unhappy with.  I haven't had time to try any of this myself, but the possibilities for combining enfuse's automated results with manual fine tuning using the old techniques are quite staggering.  Whether it's worthwhile spending all that extra time on exposure blending depends on how (un)happy you are with enfuse's automated output.  My take is that the exposure blending done by enfuse is easily good enough for production work.  YMMV.)

So play away! :)

BBB
Reply all
Reply to author
Forward
0 new messages