Automatic sky glow removal and other misc tweaks #2

merged 4 commits into from Aug 16, 2016


None yet
2 participants

berrange commented Aug 15, 2016

The main purpose of this pull request is to add support for automatic sky glow removal. It works by performing a gaussian blur on a copy of each light frame and subtracting that just after subtracting the dark frame. This almost completely eliminates the sky glow, as well as other annoyances such as clouds which may have drifted through. It is rather slow, but totally worth it for people with heavily light polluted skies as seen in this blog post

The other nice fix is to make use of high precision images in GIMP 2.9 to improve quality of the final result

berrange added some commits Aug 15, 2016

sort light/dark frame filenames before processing
The plugin config window shows the current filename being
processed in the progress display. By sorting filenames
alphabetically the user can get an idea of progress
through their batch.

Signed-off-by: Daniel P. Berrange <>
Put plugin directly in <Image>File/Create
Don't put the plug in a "Python-Fu" sub-directory as it
slows down execution by requiring the user to navigate
yet another level in the menu hierarchy. Also the choice
of programming language a plugin was written in, is pretty
much irrelevant to a user running it.

Signed-off-by: Daniel P. Berrange <>
add option to automatically remove skyglow
Light pollution skyglow can be automatically removed from light
frames with the following technique

 - Create a duplicate layer of light frame
 - Apply gaussian blur to the dup layer with huge radius
 - Subtract duplicate layer from original light frame

This technique only works nicely when the light frame does not
contain any foreground objects. If foreground objects are present
then it will create halos around their edges.

Signed-off-by: Daniel P. Berrange <>
Support high precision images
If running with GIMP 2.9 it is possible to use high precision
images (ie 16/32/64 bpp) to improve quality of the result.
Automatically create images with the same precision as the
light/dark frames being loaded. Potentially we should use
16-bit depth, even if the original image is only 8-bit since
the repeated layer operations could degrade quality.

Signed-off-by: Daniel P. Berrange <>

@themaninthesuitcase themaninthesuitcase merged commit 81cc7e5 into themaninthesuitcase:master Aug 16, 2016

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment