ESD vs Linear Fit for large image sets

JPetruzzi

Active member
Jun 18, 2020
35
9
Just wondering if anyone has done any comparisons re: ESD vs Linear Fit Clipping for sets of images > 100 frames or so...I've been playing around with a sort of poor data set but plan to do something more systematic after I get some imaging time in again. Anyone else done a comparison, or have thoughts/suggestions? (Apologies if this topic has been done before, couldn't find it) TIA
 
Last edited:

ngc1535

PTeam Member
Feb 1, 2014
532
68
AdamBlockStudios.com
HI,

So... I have read your question a few times- but still can't quite make it out.
"G"ESD is the new rejection algorithm in Image Integration and Linear Fit is a way to match images offset and signal values.
They are different kinds of tools for different purposes. Does the "ESD" refer to something else?
Thanks,
Adam
 
  • Like
Reactions: JPetruzzi

JPetruzzi

Active member
Jun 18, 2020
35
9
Sorry, meant "Linear Fit Clipping" which is also a rejection algorithm, have edited initial post to reflect. : ) 1596524995187.png
 
Last edited:

ngc1535

PTeam Member
Feb 1, 2014
532
68
AdamBlockStudios.com
Ok... so, I have given this some thought and I can only offer generalizations.
First, with a large sample (many values) such as what you suggest of more than 50 frames- I suspect that the difference between rejection algorithms (given the proper commensurate parameter choices and thresholds) will be small. Heck, the process of simply averaging large numbers of values lessens the significance of outliers to an almost rejection-like result. That being said, I did find I could occasionally get unexpected results from the Linear Fit Clipping method. This probably had to do with something systematic about the data affecting the fit and subsequent rejections. I have found GESD to be quite stable so far. I fear I am rejecting more with GESD than I normally do- so I think I need more experience with it. I haven't seen any unexpected rejections (or lack of) when using it.

-adam
 

JPetruzzi

Active member
Jun 18, 2020
35
9
Thank you for the feedback....I've been playing with an extreme case (120 x 3 sec untracked exposures of Neowise, taken just over the horizon so a really pronounced sky gradient) and, true to the algorithm descriptions, Linear Fit Clipping seems to have done a slightly better job evening the gradient but I suspect for better quality data (G)ESD is probably the more effective choice, will need to spend some time messing with parameters...as you say, need more experience with it. I think I have overall been over-rejecting...fortunately I never get tired of playing with data. : ) I hope to get some decent deep sky frames next week once the moon is gone, whatever I end up with will be my next test case. Thanks!