Highest Median noise reduction with no rejection?

[mstriebeck]

Hi,

I am stacking 60 5min frames with ImageIntegration. I am imaging from a very light polluted area. I compared the noise reduction with "no rejection", "sigma clipping", "winsorized sigma clipping" and "linear fit clipping". Here are the values that I got:

No rejection: 1.6610
Sigma Clipping: 1.5744
Winsorized Sigma Clipping: 1.6190
Linear Fit Clipping: 1.6165

I find it strange that I get the highest reduction without any pixel rejection. Especially with my conditions, I expected to get crazy outliers and much better noise reduction with pixel rejection.

Happy to make the subs or out images (rejection_high, rejection_low, integrated image) available for analysis.

Thanks for any pointers or ideas.

      MarkS

[RickS]

That's expected behaviour... you'll get maximum SNR from an average combine with no rejection.  The crazy outlier stuff is unwanted signal, not random noise.  You want to get as close to this SNR as possible by tweaking the rejection algorithm and parameters while still maintaining a clean result.

I first read about this here: http://www.astrosurf.com/jordigallego/articles/Image_integration_JGallego.ppt

Cheers,
Rick.

[mstriebeck]

Doooooh! I read his presentation but completely misunderstood this part. Thanks for clarifying!