PixInsight offers two powerful background modelization methods: AutomaticBackgroundExtractor (ABE) and DynamicBackgroundExctractor (DBE).

Many people simply use ABE and are satisfied with the results, thinking that PixInsight chooses the optimal background for you, when in fact, you can do a much better job with DBE.

So what do ABE and DBE actually do? They basically equalize the background of your image based on a model that the tools generate. In ABE, the process is automatic – you just drag the new instance icon over your image and the background gets modeled and equalized. But there’s a problem with ABE: because it is automatic, it takes a brute force approach to sampling the background of your image. While this makes the tool easy to use, it does so at the cost of perhaps coming up with a less than perfect result.

Here’s the result of running ABE on an image of comet Neowise (2020 F3):

The background of this image is very uneven: the corners exhibit significant vignetting, there’s a dark halo around the comet itself and there are structures between the comet and the edge of the image. I created this result by using ABE at its default settings.

Here’s the original image before applying ABE:

The green areas highlight darker regions and the area between the comet and the edge of the image is quite bright.

When you use ABE, it places samples on your image and uses them to model the background. At its default, ABE places samples that are sized at 5×5 pixels and separates them by 5 pixels. You can get ABE to show the samples it uses by selecting the ‘Draw sample boxes’ option, as I have done here:

When using the default options, ABE comes up with the following samples:

That’s a lot of samples, and notice how they get close to the comet – this explains the uneven background that ABE came up with.

Here’s the same image, except this time, I used DBE to manually place samples on the image:

Note that the vignetting is gone and the background around the comet is much more even. Here are the samples I placed using DBE:

Note that I resized the samples to be 30×30 pixels and I placed them at the corners of the image and between the comet and the edge of the image. The samples allowed DBE to come up with a smooth background model and it produced the picture you saw before this one.

From this, I hope to show you that although PixInsight is very powerful and is great at processing images, you are ultimately the best decision-maker about your images. You can see things that PixInsight cannot see and so can make better decisions about how to use a particular tool.

Conclusion

In this article I demonstrated the difference between using ABE and DBE on the same image. I explained what both processes do and showed you one way to use DBE to improve an image.