[Solved] Interpolation artifacts with Lanczos, Catmull-Rom Spline, Mitchell-Netravali, Cubic B-Spline

Posted by: Dong Hun

Hi Mabula

 

I'm getting little bit of interpolation artifacts for the data set that I'm working with.

There are dark pixels at the edge of some stars when I use Lanczos-3 as the interpolation algorithm.

 

In fact the artifacts are a lot worse in Pixinsight.

 

They are much less noticeable with APP.

 

If I use bilinear algorithm, the artifacts go away but the image is not as sharp as one from Lanczos-3(bigger FWHM value).

 

I'm wondering if there is a way to preserve FWHM but remove the artifacts.

 

Thank you very much.

Donghun

 

Hi Donghun,

Yes, I clearly see the artefact. This artefact is known as undershoot. This will always happen with a data interpolation algorithm like Lanczos.

But there is a solution to prevent this in APP. You will need to enable the "no under/overshoot" selectbox below the pixel interpolation filter in 6) INTEGRATE. This should prevent these dark pixels at the star borders due to undershoot.

Let me know if this solves the artefacts with your data.

Mabula

Under and Overshoot happens because a filter like Lanczos has positive and negative lobes see:

https://en.wikipedia.org/wiki/Lanczos_resampling

https://en.wikipedia.org/wiki/Ringing_artifacts