how should I make a indoor 360° panorama

157 views
Skip to first unread message

chaosjug

unread,
Dec 18, 2021, 3:15:03 AM12/18/21
to hugi...@googlegroups.com
Hi,

I'm looking for the best way to do make 360° indoor panoramas. I'm using a
panorama head and the images align well. No issues there.
What I'm wondering is this:
I'm shooting stack with different exposure. Is that the right thing? I need
different exposure when shooting against a window versus a dark part of the
room.
How should I build the final image from that? I tried the "fused" and
"blended_fused" but it is not totally clear to me what is actually the
difference. Is it the order if stacking or building the panorama is done first?
Or is there anything else to it? Which is better?

When using "blended_fused" I get three intermediate panoramas with different
exposure although there are 7 images in each stack. It is unclear to me why
that happens. I get the following warnings:
enblend: warning: some images are redundant and will not be blended
enblend: note: usually this means that at least one of the images
enblend: note: does not belong to the set

As I don't have a wide angle lens, I need a lot of images and thus it takes a
lot of time to merge. Are there any tricks to speed things up?

Regards,
Stephan



Florian Königstein

unread,
Dec 19, 2021, 11:13:05 AM12/19/21
to hugin and other free panoramic software
The stitching could be speeded up using multiblend.exe from David Horman instead of enblend, especially if many images must be stitched. As far as I know, the time for stitching with enblend increases with the square of the number of images, whereas it is proportional to this number with multiblend.

dkloi

unread,
Dec 19, 2021, 3:53:41 PM12/19/21
to hugin and other free panoramic software
From the Hugin manual at https://wiki.panotools.org/Hugin_Stitcher_tab

"If Exposure fused from any arrangement is enabled then hugin will seam blend images with similar exposure with enblend and than it will exposure fuse them using enfuse. This variant is often much more successful than Exposure fused from stacks in two situations:
  • Where entire panoramas have been shot at each EV level consecutively rather than each shot bracketed, in this case it isn't guaranteed that shots will line up into the approximate stacks expected by the Exposure fused from stacks option.
  • When the panorama has been shot entirely on automatic exposure, in this situation it is useful to seam blend adjacent photos with small EV differences, but then exposure fuse larger EV differences - As effectively happens with this option.

Note that Hugin uses a default threshold of 0.5 EV exposure difference to determine which photos can be fused into each layer. This threshold can be modified on the Photos tab in the Expert mode (selecting group by Output layers)."

What exposure bracket settings are you using? If it's 0.5EV steps, then maybe Hugin is combining 2 exposure planes?

Are you able to share your project and a set of images for diagnosis? Downsized, compressed would be fine for testing exposure fusion.

There are other ways of handling high contrast panos. I have some articles that may be useful at http://www.dkloi.co.uk/?p=1454 http://www.dkloi.co.uk/?p=1501 and http://www.dkloi.co.uk/?p=1518

chaosjug

unread,
Dec 19, 2021, 6:59:26 PM12/19/21
to hugi...@googlegroups.com
Hi,

thanks a lot for your input!
My bracket setting was 0.3EV so no wonder I was missing output layers. I had
to search a bit, because the german translation is a bit fuzzy. In german it
is named "Belichtungsebenen" which I would translate as exposure layer, so it
is not so clear that this is for the output layers.
From what I read in the articles you linked, I'm shooting way too many
pictures. I'll give it a try with less.
At the moment I shoot jpgs but I could do raw. Is there much benefit? I see
different white balance in the original images but hugin seems to handle that
just fine.

I also tried multiblend. It seems to have a bit more trouble if the
controlpoints are not perfect but it is indeed faster. That doesn't help with
nona and enfuse which took a lot of time with my big stacks but if I don't
need that, then it will speed up the process a lot.

I also found some articles about lens calibration. How do I know if I should
do that?

If anyone knows articles/tutorials about 360° and/or indoor panos please
share. I find it really hard to find something that covers more than the basics
and is not outdated (or at least old and thus unclear to me if still valid).

Am Sonntag, 19. Dezember 2021, 21:53:41 CET schrieb 'dkloi' via hugin and
other free panoramic software:
> From the Hugin manual at https://wiki.panotools.org/Hugin_Stitcher_tab
>
> "If *Exposure fused from any arrangement* is enabled then hugin will seam
> blend images with similar exposure with enblend
> <https://wiki.panotools.org/Enblend> and than it will exposure fuse
> <https://wiki.panotools.org/index.php?title=Exposure_fusion&action=edit&redl
> ink=1> them using enfuse <https://wiki.panotools.org/Enfuse>. This variant
> is often much more successful than *Exposure fused from stacks* in two
> situations:
>
> - Where entire panoramas have been shot at each EV level consecutively
> rather than each shot bracketed <https://wiki.panotools.org/Bracketing>,
> in this case it isn't guaranteed that shots will line up into the
> approximate stacks expected by the *Exposure fused from stacks* option.
>
>
> - When the panorama has been shot entirely on automatic exposure, in
> this situation it is useful to seam blend adjacent photos with small EV
> differences, but then exposure fuse larger EV differences - As
> effectively happens with this option.
>
> Note that Hugin uses a default threshold of 0.5 EV exposure difference to
> determine which photos can be fused into each layer. This threshold can be
> modified on the Photos tab <https://wiki.panotools.org/Hugin_Photos_tab> in

dkloi

unread,
Dec 20, 2021, 11:41:54 AM12/20/21
to hugin and other free panoramic software
You should definitely be shooting raw. Apart from the ability to adjust white balance, the amount of exposure latitude in the raw file is typically much greater than what is preserved in an in-camera jpg, especially if you are using a larger sensor camera e.g. DSLR or mirrorless.

In raw development, I often max out highlight recovery and use extensively shadow recovery. This can gain a lot of detail that would normally be lost in standard development.

If I bracket, it's usually in steps of 2EV. I usually don't bother with overexposing but mainly care about making sure I don't lose highlight detail, hence may shoot -2EV or -3EV in addition to the autoexposure value. But this all depends on your camera and its sensor characteristics, you may need the +EV brackets if shadow performance isn't good.

Lens calibration isn't usually required but could help in some situations where there are not enough control points to give a good lens geometry optimisation. I haven't bothered calibrating my lens, other than inputting the calibrated focal length (as the lens does not transmit metadata to the camera).

To make your life easier, I suggest getting a fisheye lens, there are quite a few affordable options. This would reduce the number of shots you need to take as well as reduce the processing time. What output size do you need and what camera are you using?

What isn't covered in the articles I mentioned that you'd like to know more about? Have you played around with the example files?

chaosjug

unread,
Dec 28, 2021, 4:37:23 PM12/28/21
to hugi...@googlegroups.com
I'll give raw a try, but I haven't used it much before and struggle to get
better results than just using the jpeg...

I have tried to capture the same room once on auto and once with bracket
(+/-1EV) to get a feeling for the difference. With auto, the white walls get
shadows. With bracket that is no issue. But in both cases the final image looks
as if taken at the same exposure all around. So dark parts are too dark and
bright parts too bright. Even if I add a specially exposed shot e.g. of a
bright window and set the masks accordingly.
Looking at your Ripon Cathedral Library example, the trick seems to be to make
two panos and use enblend again to merge them. But how do I actually create a
darker pano? And for the final enblend I probably need to pass the masks again,
right?

I guess I'll have to invest into a fisheye lens if I don't want to send hours
on setting control points ... I'm just a bit hesitant as I have a OMD E-M10
Mark II and the MFT format seems to be dying out. Which is sad, as I really
like the form factor. Much less to carry on hikes which is my main use case.

Your articles are great, because they are not a step by step guide without
explanation as so many tutorials on the basic how to stitch with hugin. But on
the other side I would need some more step by step because I can't translate
that into how to do this for real like above.
And of course: looking at how others are doing things is a great way to learn
imho.

Am Montag, 20. Dezember 2021, 17:41:53 CET schrieb 'dkloi' via hugin and other

dkloi

unread,
Jan 3, 2022, 1:14:59 PM1/3/22
to hugin and other free panoramic software
You're losing out on a lot of potential of your camera not shooting raw. In high contrast situations, this is especially the case where the ability to recover highlights and boost shadows as needed in post-processing is very advantageous, not to mention the ability to tweak white balance easily, even in extreme lighting conditions. The limited bit depth together with lossy compression of jpg images does pose significant limitations when doing panoramic photography in challenging situations.

Not sure what you mean by, " With auto, the white walls get shadows." Perhaps this is due to vignetting, lens light fall-off?

In order to tame intra-scene contrast, the trick I use is to produce 2 (or more) output versions of the same scene at different brightnesses. This is using the "Panorama preview" window and the "EV:" setting. You can output a default brightness (hit the green arrow) and output a pano, then produce other output panos that are over and under exposed to capture the shadow and highlight detail as required. Then use enfuse to exposure blend the 3 panos together, see https://i1.wp.com/www.dkloi.co.uk/wp-content/uploads/2020/01/06EnfuseExample.png from http://www.dkloi.co.uk/?p=1454

MFT fisheye lenses aren't that expensive, the Samyang 7.5mm is about £220 from Amazon. I'm not sure whether this requires 4, 5, or 6 shots around, but even a 6+Z+N pattern would save you a lot of time.

Glad you found my articles useful. They're not really aimed at the novice, the already existing step-by-step guides are probably better for those who are starting off. But this group is always a good place to ask questions about things that aren't clear.

chaosjug

unread,
Jan 3, 2022, 3:28:39 PM1/3/22
to hugi...@googlegroups.com
I think the shadows have the same pattern than the original images. So
vignetting might be an issue, but then I would expect to see the same issue
with the bracket version. It is also sometimes much less and hardly noticeable
but I'm not sure what I did. It might be, that the photometric optimisation
isn't always working as it should. That part is pretty much black magic for me
at the moment. The geometric I've mostly figured out. Only camera translations
for the nadir shot is giving me troubles but I get mostly acceptable results
without it.

So you are doing one pano in nomal EV. And then one for under or over exposure
where you might add another frame which was shot explicitly for that like in
the Ripon Cathedral Library? I wasn't expecting to find the EV correction on
what I had stamped as beginners interface. I may need to look more closely
what I can change the as well.

Am Montag, 3. Januar 2022, 19:14:59 CET schrieb 'dkloi' via hugin and other
free panoramic software:

dkloi

unread,
Jan 3, 2022, 5:46:07 PM1/3/22
to hugin and other free panoramic software
In the Ripon Cathedral example, look on the Masks tab and DSC01617.jpg, photo #3. You can see the green include masks so that when it is toggled to be included in the final pano, it will mask out the other images in those bright areas.

Then go to the "Panorama preview" and you can see what happens when you toggle on and off photo #3. If it is toggled on (i.e. all photos actives), then you can decrease the brightness until the window is not blown out, this is around EV5.97 but the rest of the pano is a bit dark. Export this to get an "underexposed" pano. Deactivate photo #3 and return the brightness of the pano to "normal" (should be EV4.97) and export this version for a default exposure. There's no need to for brighter version for this particular pano but the same would apply if there were some shadow areas you wanted to brighten up a bit. Now run enfuse on the exported dark and normal exposures to get an exposure blended version of the pano.

This one https://www.360cities.net/image/france-toulouse-notre-dame-du-taur-south-chapel required a range of exposure times from 20s to 1/15th and I exposure blended both under and overexposed exports.

A thing to note is that this technique of rendering outputs at different brightnesses doesn't require taking extra shots in some cases. It may be that the brightness in one frame/direction is much different from the brightness in another, although the intra-frame contrast may be tolerable in each shot.
Reply all
Reply to author
Forward
0 new messages