Random green pixels with Hugin 20.0 on Windows

[Paul Hardaker]

Hi everyone, I'm a relatively new user of Hugin and hoping for some expert advice on how to correct an odd problem.  I'm working on a set of fairly low-light panos (approx 10 images in full-res TIFF format for each pano) with Hugin 20.0 on Windows 10.  For several of the panos I'm doing, the resulting stitched and blended image has small areas of mainly bright green pixels appearing in areas that should be close to black (and which are black or close to black on the original images).  See attached file as an example.  Only really noticeable when fully zoomed in on the pano image.  I've tried changing various settings within Hugin but so far can't eliminate the problem.  The only setting that seems to reduce the effect to some small degree is the Photometric Optimizer setting (Preferences - Misc).  What's really odd is that I've run the same pano creation process using the exact same source images using Hugin on Linux (Fedora) and it works perfectly - no green pixels.  Other than these areas of random green (also some red) pixels in the pano images created under Win10, they come out just perfect.  Any help / advice on how to correct this annoying problem very much appreciate.

Apologies in advance if this is a question that has come up (and been answered) before.

Thanks!

[T. Modes]

Hi,

first you need to narrow down the issue. The issue can be introduced during (a) remapping by nona or (b) by blending with enblend/enfuse.

Check the intermediate (remapped) images - you can activate them on the stitcher tab.

If the remapped images show already the green spots (case (a)), you could try to reset the response curve and the color balance values of the images.

If the remapped images don't show the green spots, then they are a result of the blending step (case (b)). In this case you could try to add the switch --blend-colorspace=identity to the enblend parameters on the stitcher tab.

[Paul Hardaker]

Many thanks for this advice.  Intermediate images are clear (no green/red pixel "pollution").  I added the switch  --blend-colorspace=identity to enblend as suggested, and the problem is now solved!  Perfectly clean pano image.

Odd that the Linux version of Hugin does not seem to have this issue.  Some subtle difference in the underlying libraries between the difference OS platforms maybe.

Thanks again.

[T. Modes]

Hi,

good to hear.

pj.ha...@gmail.com schrieb am Dienstag, 4. Mai 2021 um 23:38:27 UTC+2:

Odd that the Linux version of Hugin does not seem to have this issue.  Some subtle difference in the underlying libraries between the difference OS platforms maybe.

What is the output of

enblend --show-software-components

and

enblend --version --verbose

on Linux?

This should print some of the used libraries and the version/compiler used.  (For the Windows version I can do it for myself.)

Thanks

[Paul Hardaker]

$ uname -a

Linux localhost.localdomain 5.11.15-200.fc33.x86_64 #1 SMP Fri Apr 16 13:41:20 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux

$ enblend --show-software-components

Compiler

  g++ 10.2.1

  implementing OpenMP standard of 2015-11

Libraries

  GSL:        2.6

  Little CMS: 2.10.0

  Vigra:      1.11.1

$ enblend --version --verbose

enblend 4.2

Extra feature: image cache: no

Extra feature: OpenMP: yes

  - version 2015-11

  - support for dynamic adjustment of the number of threads;

    dynamic adjustment disabled by default

  - using 4 processors and up to 4 threads

  - allocating thread-local dynamic memory with malloc/free

Extra feature: OpenCL: no

Copyright (C) 2004-2009 Andrew Mihal.

Copyright (C) 2009-2016 Christoph Spiel.

License GPLv2+: GNU GPL version 2 or later <http://www.gnu.org/licenses/gpl.html>

This is free software: you are free to change and redistribute it.

There is NO WARRANTY, to the extent permitted by law.

Written by Andrew Mihal, Christoph Spiel and others.

[T. Modes]

Thanks.

It appears there updates to several libs. So it not so easy to isolate the necessary change.