Strange Hugin failure

[Greg 'groggy' Lehey]

A friend of mine recently published a panorama that he had stitched
from first principles with Mathematica. It didn't look bad. But I
thought it could be done better with Hugin. I was wrong.

First, I ran it through my scripts, which effectively run pto_gen,
cpfind, celeste_standalone, cpclean and autooptimiser. The result
showed very uneven lighting.

OK, I thought that maybe I had something in my scripts that wasn't
doing the right thing, so I tried running it in a vanilla version of
Hugin without any ~/.hugin file. Things were *much* worse. Hugin
couldn't align the images at all. It seems that it couldn't
understand the exposure info, and it made the component images
progressively darker.

It also came with a popup "The project covers a big brightness
range,". But that's not what the images show. They cover a range 9.8
to 11.4 EV, and that matches the lighting. About the only thing
that's unusual is that the photos, taken with a Pixel 8 Pro, were
taken at a sensitivity of only 15 ISO. But that shouldn't make any
difference.

I've tried this with the latest version of Hugin and also with a 5
year old one, and the results are the same. You can see the summary,
with all the images above and more, at
http://www.lemis.com/grog/diary-jun2024.php#D-20240611-005714 , which
includes the location of the images (about 400 MB).

Any ideas?

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

[T. Modes]

Hi Greg,

Groogle schrieb am Mittwoch, 12. Juni 2024 um 07:17:06 UTC+2:

It also came with a popup "The project covers a big brightness
range,". But that's not what the images show. They cover a range 9.8
to 11.4 EV, and that matches the lighting. About the only thing
that's unusual is that the photos, taken with a Pixel 8 Pro, were
taken at a sensitivity of only 15 ISO. But that shouldn't make any
difference.

The EXIF data say a brightness value between 7.1 and 9.16. From the exposure time, iso and aperture exiftool and Hugin calculates light values between 12.1 and 14.2.

So more than 2 stops difference -> that triggers the warning.

exiftool -T -Filename -ExposureTime -ISO -BrightnessValue -LightValue .
Filename ExposureTime ISO BrightnessValue LightValue
PXL_20240608_180517602.jpg 1/351 15 9.16 14.2
PXL_20240608_180522815.jpg 1/338 15 9.11 14.1
PXL_20240608_180529707.jpg 1/291 14 8.99 14.0
PXL_20240608_180536796.jpg 1/291 14 8.99 14.0
PXL_20240608_180544529.jpg 1/287 14 8.97 14.0
PXL_20240608_180555179.jpg 1/274 15 8.81 13.8
PXL_20240608_180603306.jpg 1/253 15 8.69 13.7
PXL_20240608_180610037.jpg 1/166 15 8.08 13.1
PXL_20240608_180616602.jpg 1/204 19 8.04 13.0
PXL_20240608_180626057.jpg 1/175 16 8.07 13.1
PXL_20240608_180631289.jpg 1/175 15 8.16 13.2
PXL_20240608_180646882.jpg 1/184 15 8.23 13.2
PXL_20240608_180706253.jpg 1/179 15 8.19 13.2
PXL_20240608_180717793.jpg 1/182 15 8.22 13.2
PXL_20240608_180723986.jpg 1/159 16 7.93 12.9
PXL_20240608_180734844.jpg 1/172 15 8.14 13.1
PXL_20240608_180742451.jpg 1/168 17 7.92 12.9
PXL_20240608_180749280.jpg 1/160 17 7.85 12.9
PXL_20240608_180759125.jpg 1/149 15 7.93 12.9
PXL_20240608_180813662.jpg 1/114 15 7.54 12.5
PXL_20240608_180819953.jpg 1/117 16 7.48 12.5
PXL_20240608_180832181.jpg 1/114 15 7.54 12.5
PXL_20240608_180846659.jpg 1/119 15 7.6 12.6
PXL_20240608_180908776.jpg 1/84 15 7.1 12.1
PXL_20240608_180925043.jpg 1/85 15 7.12 12.1
PXL_20240608_180931964.jpg 1/85 15 7.12 12.1
PXL_20240608_180947768.jpg 1/120 15 7.61 12.6
PXL_20240608_180953798.jpg 1/136 16 7.7 12.7
PXL_20240608_181000599.jpg 1/124 15 7.66 12.7
PXL_20240608_181007744.jpg 1/113 15 7.53 12.5

Maybe the issue for the big difference is that the images have been heavily processed on the iPhone:

[IFD0]          Software                        : HDR+ 1.0.621982163nd
[ExifIFD]       Composite Image                 : Composite Image Captured While Shooting

So the image brightness does not match a single exposure (and therefore the EXIF information). This could explain the differences after loading.

 

I've tried this with the latest version of Hugin and also with a 5
year old one, and the results are the same. You can see the summary,
with all the images above and more, at

Not sure what goes wrong on your side. But the downsized results with assistant from default branch is attached. This looks good for me.

Now the light value ranges from 13.6 to 14.2.

The default branch contains a bug fix regarding loading images from the iPhone taken with digital zoom.

Apple does not follow the EXIF standard regarding the digital zoom. Therefore in older version the images were read with an too high HFOV (in your case this should only make a factor of 1.1).

Thomas

[Claudio Rocha]

I can't reprodce the error. The images blend just nice

Hugin 2021.0.0.52df0f76c700 on Linux

enblend-enfuse
Version: 4.2-10

[Greg 'groggy' Lehey]

On Friday, 14 June 2024 at 15:56:48 -1000, David W. Jones wrote:
> On 6/14/24 15:49, David W. Jones wrote:
>> Here's what I got after running the tapiola images through Hugin
>> Pre-Release 2023.0.0.548f2a905b6a using the Assistant. It looks like
>> there might be unnecessary images, maybe somewhere around 11, 12, 13,
>> 14. There were a lot of bad control points in that area. But I don't
>> think the final blend shows any big exposure changes.
>
> Well, got a message from Google Groups saying it has permanently
> removed my "illegal content," so I guess I won't bother. The
> generated image is 1.6MB, the PTO much smaller, let me know if
> you're interested.

Thanks. It got through to me fine. I'm still scratching my head
about what happened, but I'll get back to you and Thomas.

> Stupid Google.

It's not picking on you. My first attempt at my message was also
rejected:
http://www.lemis.com/grog/diary-jun2024.php?subtitle=Google:%20You%20have%20violated%20our%20policy&article=D-20240613-021154#D-20240613-021154
I was particularly annoyed because they give me no choice of getting a
Real Human to look at the issue. My guess is that they didn't like
the URLs I included in the first message.

Greg
--
Sent from my desktop computer.

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

[David W. Jones]

On 6/14/24 16:38, Greg 'groggy' Lehey wrote:

On Friday, 14 June 2024 at 15:56:48 -1000, David W. Jones wrote:

On 6/14/24 15:49, David W. Jones wrote:

Here's what I got after running the tapiola images through Hugin
Pre-Release 2023.0.0.548f2a905b6a using the Assistant. It looks like
there might be unnecessary images, maybe somewhere around 11, 12, 13,
14. There were a lot of bad control points in that area. But I don't
think the final blend shows any big exposure changes.

Well, got a message from Google Groups saying it has permanently
removed my "illegal content," so I guess I won't bother. The
generated image is 1.6MB, the PTO much smaller, let me know if
you're interested.

Thanks.  It got through to me fine.  I'm still scratching my head
about what happened, but I'll get back to you and Thomas.

Thanks.

I also generated a full-sized version of it (998MB TIF) and that looks fine. It's a nice image, good shadows and contrast in Luminance HDR's Ferwerda HDR mapping.

Stupid Google.

It's not picking on you.  My first attempt at my message was also
rejected:
http://www.lemis.com/grog/diary-jun2024.php?subtitle=Google:%20You%20have%20violated%20our%20policy&article=D-20240613-021154#D-20240613-021154
I was particularly annoyed because they give me no choice of getting a
Real Human to look at the issue.  My guess is that they didn't like
the URLs I included in the first message.

Greg

Didn't you hear? Google's in the AI business! I don't think they have Real Humans™ there anymore.

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