I’d read a few bits and pieces online about Starnet++ – a software module that uses a neural network to identify and remove stars from astronomical images in order to enhance nebulosity and process it separately to stars.
The software is free and you can find it at the link below. It was originally published as a Pixinsight module, but can now be downloaded and run standalone on Windows: https://sourceforge.net/projects/starnet/
I’ve found it’s pretty good- I’ve been playing with it today on an image of the Western Veil I took a few weeks back. This is my original processing of the image:
To use the technique I started again with the stacked file and used Pixinsight to remove the background light pollution gradients, calibrate the colours and do an initial stretch. I then put the image through the Starnet routine and it returned me the image below:
I then used the Clone Stamp tool to clean it all up (possibly more time needed on this!) and tweaked the curves to give it some contrast and got this:
I really like this, but I felt it would be better with some stars blended back in, so I went back to the image I submitted and processed purely to get the brightest stars at a prominence that I liked. I then blended the two images using Pixelmath (in the way I used it here, it’s identical to blending layers in Photoshop or GIMP with lighten):