Experience with 7Artisans 4 mm fisheye?

[Greg 'groggy' Lehey]

I have recently got a 7Artisans 4 mm f/2.8 circular fisheye lens with
an angle of view of claimed 225°. I'm using it on an Olympus E-PM2,
and I'm having no luck at all stitching panoramas. I suspect that
there is some issue with Hugin misinterpreting the Exif data, but
before I go to even more work, a question: is anybody using a lens of
this nature? Are there any tricks?

I'll follow up with more details if nobody sees an immediate issue.
You can also take a look at the three images
http://www.lemis.com/grog/Day/20250131/Office-0.jpeg
http://www.lemis.com/grog/Day/20250131/Office-1.jpeg
http://www.lemis.com/grog/Day/20250131/Office-2.jpeg (each about 900
kB), or at my diary at
http://www.lemis.com/grog/diary-feb2025.php?topics=p&subtitle=More%20fisheye%20fun&article=D-20250202-012736#D-20250202-012736

One thing that is interesting: the "focal length" popup sometimes
reports a horizontal angle of view of 248° (close enough, since it's
way off the end of the image), but at other times 186°, which is
clearly wrong. Is there any way to tell the GUI or the underlying
scripts what the real angle of view is?

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

[David W. Jones]

On 2/3/25 16:16, Greg 'groggy' Lehey wrote:
> I have recently got a 7Artisans 4 mm f/2.8 circular fisheye lens with
> an angle of view of claimed 225°. I'm using it on an Olympus E-PM2,
> and I'm having no luck at all stitching panoramas. I suspect that
> there is some issue with Hugin misinterpreting the Exif data, but
> before I go to even more work, a question: is anybody using a lens of
> this nature? Are there any tricks?
>
> I'll follow up with more details if nobody sees an immediate issue.
> You can also take a look at the three images
> http://www.lemis.com/grog/Day/20250131/Office-0.jpeg
> http://www.lemis.com/grog/Day/20250131/Office-1.jpeg
> http://www.lemis.com/grog/Day/20250131/Office-2.jpeg (each about 900
> kB), or at my diary at
> http://www.lemis.com/grog/diary-feb2025.php?topics=p&subtitle=More%20fisheye%20fun&article=D-20250202-012736#D-20250202-012736
>
> One thing that is interesting: the "focal length" popup sometimes
> reports a horizontal angle of view of 248° (close enough, since it's
> way off the end of the image), but at other times 186°, which is
> clearly wrong. Is there any way to tell the GUI or the underlying
> scripts what the real angle of view is?
>
> Greg

Hi, Greg!

I tried several different ways of generating control points with your
JPEGs, and was unable to generate any control points connecting any images.

I manually added some, but alignment didn't give me anything except a
jumble of spikes. The nearest I got to anything coherent was using the
Lambert Cylindrical Equal Area projection, and that just gave me a row
of separate hump-shaped images that didn't really connect.

Sorry, I have no other ideas. I don't know if everything's being
affected by the pronounced chromatic distortion around the edges of each
image, but I know nothing about using fisheye lenses like yours!

--
David W. Jones
gnome...@gmail.com
wandering the landscape of god
http://dancingtreefrog.com
My password is the last 8 digits of π.

[Greg 'groggy' Lehey]

On Monday, 3 February 2025 at 20:08:38 -1000, David W. Jones wrote:
> On 2/3/25 16:16, Greg 'groggy' Lehey wrote:
>> I have recently got a 7Artisans 4 mm f/2.8 circular fisheye lens with
>> an angle of view of claimed 225°. I'm using it on an Olympus E-PM2,
>> and I'm having no luck at all stitching panoramas. I suspect that
>> there is some issue with Hugin misinterpreting the Exif data, but
>> before I go to even more work, a question: is anybody using a lens of
>> this nature? Are there any tricks?
>>

>> One thing that is interesting: the "focal length" popup sometimes
>> reports a horizontal angle of view of 248° (close enough, since it's
>> way off the end of the image), but at other times 186°, which is
>> clearly wrong. Is there any way to tell the GUI or the underlying
>> scripts what the real angle of view is?
>

> I tried several different ways of generating control points with
> your JPEGs, and was unable to generate any control points connecting
> any images.

I think that's actually less success than I had. I did get some
control points, but there were other issues before that. In
particular, the second image was always masked in a "fleur-de-lis"
pattern (see first attachment). It didn't make any difference which
image I loaded second, and if I loaded the same image again it was
correct. See the first attached image.

> I manually added some, but alignment didn't give me anything except
> a jumble of spikes.

Did you set the lens type to circular fisheye? What angle of view did
the GUI claim? I did get something like a panorama, but it was
useless (second image).

> Sorry, I have no other ideas. I don't know if everything's being
> affected by the pronounced chromatic distortion around the edges of
> each image, but I know nothing about using fisheye lenses like
> yours!

Thanks for the try. I was really hoping to find somebody who had been
there before, and maybe somebody will pop up. FWIW I regularly make
panos with an 8 mm full frame fisheye, and I don't have any issues
there. I'm beginning to think that this might be a bug.

Greg
--
Sent from my desktop computer.

Finger groo...@gmail.com for PGP public key.

[David W. Jones]

I didn't change the lens type, I just left it to whatever it
automatically selected. I've never used a fisheye lens at all, so I have
no idea what settings or such might be needed.

I don't remember the angle of view it claimed. It reported the 4mm focal
length and about a 21x crop factor.

>> Sorry, I have no other ideas. I don't know if everything's being
>> affected by the pronounced chromatic distortion around the edges of
>> each image, but I know nothing about using fisheye lenses like
>> yours!
> Thanks for the try. I was really hoping to find somebody who had been
> there before, and maybe somebody will pop up. FWIW I regularly make
> panos with an 8 mm full frame fisheye, and I don't have any issues
> there. I'm beginning to think that this might be a bug.

Only thing that occurred to me that might be a bug is that I was
expecting a lot of barrel distortion correction, just based on how the
images looked and the wide field of view. I don't remember seeing much
correction for that from Hugin.

Hopefully, someone that uses these kinds of lenses and knows what
they're doing will chime in. I'd think that a 235deg field of view would
mean you don't need three images, just two, to cover 360deg, but what do
I know?

[Greg 'groggy' Lehey]

On Monday, 3 February 2025 at 20:43:18 -1000, David W. Jones wrote:
> On 2/3/25 20:24, Greg 'groggy' Lehey wrote:
>> On Monday, 3 February 2025 at 20:08:38 -1000, David W. Jones wrote:
>>> I manually added some, but alignment didn't give me anything except
>>> a jumble of spikes.
>>
>> Did you set the lens type to circular fisheye? What angle of view did
>> the GUI claim? I did get something like a panorama, but it was
>> useless (second image).
>
> I didn't change the lens type, I just left it to whatever it
> automatically selected. I've never used a fisheye lens at all, so I
> have no idea what settings or such might be needed.

That would explain it. As far as I know, Exif doesn't support
specifying lens projection, so if Hugin doens't know the lens, it
assumes rectilinear. You need to select the lens type manually, even
if you have Exif data.

> I don't remember the angle of view it claimed. It reported the 4mm focal
> length and about a 21x crop factor.

The crop factor would be 2x (MFT).

> Hopefully, someone that uses these kinds of lenses and knows what they're
> doing will chime in. I'd think that a 235deg field of view would mean you
> don't need three images, just two, to cover 360deg, but what do I know?

Maybe. I was thinking that 3 would make it easier to find control
points. First I need to get it to work at all.

What I have found is that the strange fleur-de-lis artefacts don't
occur if I tell Hugin that the lens is 8 mm. I might even be able to
get some useful control points, but of course the pano wouldn't close.

[Marius Loots]

I have the Laowa 4mm f/2.8 Fisheye lens. Used on Olympus OMD-EM5 II.
Not any useful results. Certainly nothing like the examples here. This
was a few years ago, and I should revisit the lens again.

Groetnis
Marius
mailto:marius...@up.ac.za
BMedSci (Hons) UP, PGCHE (UP) HCM (FPD)

add some chaos to your life and put the world in order

> --
> 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 visit https://groups.google.com/d/msgid/hugin-ptx/Z6GyqJTTG7KdGeFG%40hydra.lemis.com.

--
This message and attachments are subject to a disclaimer.

Please refer to 
http://upnet.up.ac.za/services/it/documentation/docs/004167.pdf
<http://upnet.up.ac.za/services/it/documentation/docs/004167.pdf> for
full
details.

[Bruno Postle]

Hi Greg, these pictures stitch ok.

..but you really need to centre the lens on the tripod as there is a lot of parallax error in this scene. In this situation you should only use control points on the furthest features, and ideally only features that are near the likely stitch seam (automating this would be a useful Hugin feature).

The photos are circular fisheye, so you need to check that there is a circular crop on each of the photos, this is the cause of the 'fleur-de-lis' artefact.

For a super-wide fisheye like this you probably need to optimise all lens parameters, especially d & e. This will be much more reliable if you get the parallax right.

Attached, a PTO project and stitched output, hope this helps.

--

Bruno

[Bruno Postle]

Also, once you have calibrated your distortion and vignetting parameters (use four shots for this), you should be able to take good two-shot 360 panoramas with this setup (but always optimise FoV even if you have 'good' saved lens parameters).

--

Bruno

[Bruno Postle]

As a first approximation, the point you need to rotate the lens around is where the aperture *appears* to be when you look at the lens from the outside.

With a fisheye like this, this point will vary depending on where you are looking from. So you need to look from the direction where the seam is likely to be: if you plan to take three shot panoramas, look at 60 degrees from the axis of the lens; if you plan two shot panoramas, look for the aperture from the side.

-- 

Bruno

[Greg 'groggy' Lehey]

On Tuesday, 4 February 2025 at 9:52:08 +0000, Bruno Postle wrote:
> Hi Greg, these pictures stitch ok.

Thanks! But your reply poses more questions than it answers.

> ..but you really need to centre the lens on the tripod as there is a
> lot of parallax error in this scene.

Yes. This was a first attempt, and I wasn't expecting a perfect
stitch. But I also wasn't expecting the problems I had.

> In this situation you should only use control points on the furthest
> features, and ideally only features that are near the likely stitch
> seam (automating this would be a useful Hugin feature).

Does this mean that you set the control points manually?

> The photos are circular fisheye, so you need to check that there is
> a circular crop on each of the photos,

How do I do that? You've seen what the images look like; should
anything else be done?

> this is the cause of the 'fleur-de-lis' artefact.

My guess is that you loaded the images differently. I've been doing
it via the fast panorama preview, and it was always only the second
image that was affected in this way. It doesn't happen if I load from
the main window, and the display of the fleur-de-lis pattern depends
on where the cursor is positioned in the fast panorama preview window.

> Attached, a PTO project and stitched output, hope this helps.

Thanks. Yes, this works for me, but I still don't know how to repeat
it. I've tried using the main window, and I discover that Hugin
decides that it's a circular fisheye—is this heuristics?

But it also changes the focal length from 4 mm to 2.903 mm, which
sounds like a bug. That means that if I manually enter the focal
length (from versions without Exif data, for example), it gets
incorrect information. At 4 mm the calculated angle is 185.92°, and
at 2.903 mm it's 256.18°. According to my angle of view program, the
values should be 247.88° and 341.45°, in each case obviously including
the black borders.

So: I think we have at least two bugs here, the fleur-de-lis pattern
for the second image in the fast panorama preview and the incorrect
angular calculation.

[Bruno Postle]

On Wed, 5 Feb 2025, 03:45 Greg 'groggy' Lehey wrote:

Thanks!  But your reply poses more questions than it answers.

Questions are good!

> lot of parallax error in this scene.

> In this situation you should only use control points on the furthest
> features, and ideally only features that are near the likely stitch
> seam (automating this would be a useful Hugin feature).

Does this mean that you set the control points manually?

Some manual control points were needed. The main thing was to delete control points on foreground objects, and background points away from the likely seam positions.

> The photos are circular fisheye, so you need to check that there is
> a circular crop on each of the photos,

How do I do that?  You've seen what the images look like; should
anything else be done?

In the mask/crop tab, make sure there is a visible 'crop circle' for each image, there is also a checkbox used to link the crop for all images.

> this is the cause of the 'fleur-de-lis' artefact.

My guess is that you loaded the images differently.  I've been doing
it via the fast panorama preview

I didn't use the preview, but the functions are the same. Once you have this rig calibrated in the lens database, you should be able to do everything from the preview.

Thanks.  Yes, this works for me, but I still don't know how to repeat
it.  I've tried using the main window, and I discover that Hugin
decides that it's a circular fisheye—is this heuristics?

Yes, Hugin stores lens settings and uses them next time.

But it also changes the focal length from 4 mm to 2.903 mm, which
sounds like a bug.  That means that if I manually enter the focal
length (from versions without Exif data, for example), it gets
incorrect information.  At 4 mm the calculated angle is 185.92°, and
at 2.903 mm it's 256.18°.

The stitch only uses the central 120 degrees of each photo. Control points in the periphery, ie. areas that are not contributing to the final image, are counter-productive (because of the large parallax error).

So, in this scene, the angle of view could be wildly off and it wouldn't make any difference - in fact letting it go wild probably helped lining up the bits in the centre we do care about.

When you do calibrate this lens, using four photos and rotating around the no-parallax point, you can ensure there are control points in the periphery, and you will get an accurate angle of view.

Also, when you calibrate the lens, there are other input fisheye types that might be a better fit for this lens, try them and see.

So: I think we have at least two bugs here, the fleur-de-lis pattern
for the second image in the fast panorama preview and the incorrect
angular calculation.

Hugin doesn't set a default crop circle when you pick 'circular fisheye'. Potentially it could guess this or even try and figure it out from the image itself, but for now you have to tell it where you want it.

-- 

Bruno

[SakiPapa]

Hello and Konnichiwa,

There are several types of fisheye lenses depending on the projection method.

In Hugin, besides the Circular fisheye, Stereographic, Equisolid and Orthographic are also fisheye lens projection methods.
You can manually change to one of these various fisheye lens projection methods.

The Wikipedia article on fisheye lenses shows the different projection methods of these fisheye lenses, as well as some examples of typical photographic lenses.
https://en.wikipedia.org/wiki/Fisheye_lens

I do not know what type 7Artisans 4 mm fisheye is, and many manufacturers do not reveal the lens projection. The exception is the Nikon OP-Fisheye. This is a special application and there should be few other examples.

In the case of circumferential fisheye lenses, I believe you may select whether to mask or crop the black area around the lens using the mask/crop tab of Hugin.

Tadahisa Sugisaki