TheSkySearchers

PI has just released a new Gradient Removal Tool. Thought I'd kick off a thread where we can share experiences. Unlike DBE, the tool works without the need to drop a bunch of sample points. There is some tuning that may need to be done but it seems pretty straight forward. I'll be curious what sorts of results people get. I've given it a whirl on a couple of galaxy images that I took from home and had some nasty and fairly complex gradients. It seemed to handle these well. Next up I' think I'm going try throwing it against a dusty target.

If this tool works as well as it seems so far, it's definitely going to make life easier once you master the different ways to tune it.

--Steve

I gave it a try yesterday on several different types of images and this is what I found.
Like GraxPert or ABE, the default settings on Galaxies, planetary nebulae or small contain nebulae work really well. But anytime there is a lot of diffuse nebulae filling a large portion of the sensor, these automated processes/scripts fail by removing too much of the nebulae.
Shawn at VisibleDark on YT just put out a video a couple of days ago showing it off on M51 and then a larger nebula. Same results. Worked fantastic on the galaxy but destroyed lots of the nebula with the default settings. At least this tool does have several parameters that can be adjusted and it will take some tinkering to figure out what works best. GraxPert has many of these settings too if you open up the stand-alone app. I would not recommend the GraxPert script built into PixInsight if you are working on a large diffuse nebula though.
Still, nothing beats DBE if you take the time to set proper points and make adjustments to the default settings.
Here's Shawn's video - https://youtu.be/pVxTL5mJ_N4?si=sxsBzfZKfD13GrjM
If you watch this video, when using GraXpert, he selects "Show BG Model". Although he only briefly shows it, you can pause the video when the background extraction model is full screen and see how much of the "good stuff" it removes.

Here's another evaluation video :



Graeme

A friend of mine just showed what GradientCorrection did for his problem with lens flares on this subject. Not my data.
Before

After

So I've been playing with this more. Similar to what others are finding, it works well on galaxies where there's not a lot of background dust. I tried it on a real dusty field around NGC 1333 and wasn't super happy with the results. It seemed to be brightening some of the non-dusty background blowing out the contrast. The image was flatter, but that's not what I needed. DBE definitely worked better for the dusty field. There may be a way to tweak the parameters to get it closer to the DBE results, but I haven't hit on the magic combo yet. I really wish it generated a background model so you could see what it was trying to do.

Just noticed the PI team released a new version of the gradient removal tool today. Incudes a model generation feature as well as a simplified model for gradient removal. The latter is supposedly similar to ABE. Think I'll try bouncing this off the March Astro Processing challenge data and see how it goes.

--Steve