Question to Resampling in PI

Uwe Deutermann:
I always thought that after a 3x Drizzle stack (in DSS) and some work on the image reducing it to a smaller size would give sharper images. I did now a comparison of a very cropped M57 image, and I see the following:https://www.astrobin.com/35txmy/?nc=user
As mentioned in the image, I kind of understand the result, but then I ask myself for what the resample tool is good? If I resize in IrfanView I do not see any major differences.
Can somebody explain the Resample process in PI and maybe more important, when to use it? I might have screwed up some of my images and did not even realize.

Hi Uwe, I  kind of agree with you. I already resampled some of the few drizzle integrations I made and there is clearly a degradation of the resolution. Kind of asked myself the same question, what is the point if you resample and obtain the same aspect as a normal integration? I experimented with drizzle 2x but abandonned afterwards. The key is probably the interpolation process and settings. I cannot be of great help I'm afraid.

Uwe Deutermann:
I always thought that after a 3x Drizzle stack (in DSS) and some work on the image reducing it to a smaller size would give sharper images. I did now a comparison of a very cropped M57 image, and I see the following:https://www.astrobin.com/35txmy/?nc=user As mentioned in the image, I kind of understand the result, but then I ask myself for what the resample tool is good? If I resize in IrfanView I do not see any major differences. Can somebody explain the Resample process in PI and maybe more important, when to use it? I might have screwed up some of my images and did not even realize.

Resample with <100% (reduce size) can be used to: reduce an image after a drizzle stack, reducing the image size for sharing on the web, or to increase SNR of oversampled images at the cost of resolution. Resample with more than 100% (enlarge size) is not used very often, at least by myself, except maybe to upscale 2x2 binned CCD data to match a 1x1 luminance dataset, but the drizzle process is more accurate for upscaling of raw data.

Bogdan Borz:
Hi Uwe, I  kind of agree with you. I already resampled some of the few drizzle integrations I made and there is clearly a degradation of the resolution. Kind of asked myself the same question, what is the point if you resample and obtain the same aspect as a normal integration? I experimented with drizzle 2x but abandonned afterwards. The key is probably the interpolation process and settings. I cannot be of great help I'm afraid.

If you stack with e.g. 2X drizzle, then resample to 50%, you will obtain the original resolution but with rounder/smoother stars (less aliasing). This only works well if the original data is undersampled, since the drizzle process was written for undersampled Hubble images.

Thank you all for the information, I think it goes along with that from what I have heard from others as well. A friend passed me this link: https://astronomy.tools/calculators/ccd_suitability which shows when under- or oversampling is happening. So for my setup I really do not need it, I still think that it gives a slightly better image when I drizzle 3x (but that might be just my imagination!).
I definitely agree with David that the Resampling is with Auto configuration not working well, I will try some others! And now that I know what can happen I will have an eye on this and will resize with other tools. Or we shall trust the astronomy tools results and just do a 1x drizzle!
Again, thank you for the discussion, this is what Astrobin is all about!