Inaccuracy in Star Roundness Classification

Hi all - I’m having some issues with filtering my frames prior to stacking. I generally set cutoffs for roundness and FWHM and that has yielded good results in the past. However, the last couple of times I have attempted to stack I have seen extremely high roundness values that do not reflect the actual quality of the frame. I’m seeing numbers on the order of 0.93 or 0.94 for frames where the stars are of objectively poor quality. Here is an Imgur link with a few examples: Siril Roundness - Album on Imgur. Any ideas as to what might be going on? I’ve just updated to 1.0.3 and the issue persists.

Hello, in your case it is likely that you have stars with two cores rather than a single star that makes the trail.
This is the main reason why it looks like a good image while it is obviously not.

That makes sense, thanks. I will run some of the frames through ASIStudio again and check to see how many stars are being detected. I can probably use ASIStudio to remove the worst frames and then filter through the remaining ones in Siril.

You can use Siril to sort out worst frames too. Using the frame selector.