All the male faces got “stubble” added.
Yes, a three-day beard is fashionable nowadays, but not necessarily back then.
1 Like
The problem is that with this jpeg the face is so noisy that it is hardly recognizable and accordingly it is difficult to get a good result. You should try this with original scan, or scan the photo again with better resolution:
Also, you can try other models and then combine them. Here is for example GFPGAN v1.4:
5 Likes
I didn’t intend to sound ungrateful. This software is a jump 10 steps forward.
I use ChatGPT almost daily for text based work, but it does not impress me as much as what you’ve shown me here re image processing.
No problem! My replay was to @Thomas_Do 
Going from that jpg to an image that looks that good is impressive.
1 Like
Having used the GIMP plug-in, Col Cast Reduction, I applied the G’MIC MS NL-Means C Noise2 Filter (by Iain Ferguson) at it’s default values, without any attempt to optimise it. Here is the result.
2 Likes
David, this afternoon I tinkered with your “cast” plugin, just using default settings. Got some good results. Is there a tip sheet or tutorial somewhere?
okieman, I hadn’t written any suggestions on the use of the plugin, because it was a modification of a previous one produced by Krikor (a “resident” of Gimp-Forum.net and the late GimpLearn forum).
The only adjustments after it has run are:
-
Change in contrast.
The Contrast layer opacity is preset at 50% but can be changed. This layer is there to try and compensate for the inevitable loss of contrast. -
Fine tune colour correction.
Using Colours/Saturation on the Colour layer will permit changing the correction from the original colour cast to it’s complement.
It is frequently worthwhile to use Colours/Auto/ Stretch Contrast before running the plugin. In fact it can do no harm, as if the contrast of the original image already exceeds approximately 80% of the histogram, no change will result.
I suggest that, after running the plugin and making adjustments to contrast and colour, a New from Visible will provide the basis for further changes.
The plugin script is heavily commented to explain it’s function. I am always open to suggestion for improvement, but the credit for the brilliant idea of pixelizing a layer to the layer dimensions (making the layer into one pixel, in order to obtain the average colour) must go to Krikor.
1 Like
Thanks for all the info!
I use VS Code for web development, so I can also peek at the comments.
…and a wave to Krikor!
1 Like
You have made me think about this plugin again!
I have now updated it with an option to stretch the contrast of the original image before applying the colour correction. The previous adjustments to contrast and colour still apply.
I have attached the updated version which I hope you will try and give me feedback as to it’s usefulness for real-world files.
The code of the plugin should be viewable using any text editor.
col-cast-reduction1.zip (3.1 KB)
Edit: I discovered an error in the creation of the Contrast layer. This has now been corrected.
Edit 2: I found that the pixelize process took forever for very large images, so this is now done on a very small layer which is sampled to provide the fill for the correction layer.
1 Like
I will give it a spin!
@patdavid
Does this look like something that could be included under the software section?
I would say yes. However, I’m a sporadic participant here.
Sure does - I’ll see about adding it.
2 Likes
@s7habo, would you mind giving a little more detail on how you used chainner for face restauration? I want to restore a really bad scan (e.g. lots of banding, low dynamic range) of a b/w photograph of the parents of a friend, which also has scratches and other issues. So far, all my attempts failed, but maybe with chainner … Unfortunately, the whole internet[TM] only knows chainner for ai upscaling. Any pointer welcome.
Hm, I tried upscaling with a scale of 1 and used the gfpgan 1.4 model and it worked very well. So, sorry for the noise, I figured it out by myself 
.
1 Like
Unless I’m mistaken, here’s a download link (please read their docs before jumping in) …
GFPGAN
Step-by-step …
GFPGAN on MakeUseOf
One of several videos on YouTube…
GFPGAN restoration video
1 Like