Sunday, September 13, 2015

Noise Reduction in Pixinsight: TGVDenoise vs. MultiscaleLinearTransform

In my last images, I started using TGVDenoise last year. This was based on the easy-of-use but also some analysis about its effectiveness. But I often had the problem that it created blobs on the image that were hard to fix.

Here is an example:


After TGVDenoise

I tried a lot, but could never get a smooth background. These blobs can be minimized using the ACDNR process after stretching the image, but I found that too cumbersome and didn't want to make too many changes to the image.

Kayron Mercieca recently published a comparison of the different noise reduction processes and their pro's and con's. Based on the great explanations there, I tried to use MultiscaleLinearTransform:

This looks much better - and thanks to the great explanation in the blog post I even understand now how it works and how I need to adjust it.

So, I guess for now, I'll keep using MultiscaleLinearTransform.