SubframeSelection question re: SNR weight

[MineralMike]

I very much like the new SubframeSelector process, and have a question about the term "SNR weight."

Why is the word "weight" present? How does "SNR weight" differ from plain old "SNR?"

Just wondering.... Thanks.

--- mike

[pfile]

i believe this doc refers to the script, but very likely the meaning of the keyword is the same between the two versions of SFS:

https://pixinsight.com/doc/scripts/SubframeSelector/SubframeSelector.html#subframe_property_SNRWeight

rob

[aworonow]

Unless things have changed since Juan wrote (https://pixinsight.com/forum/index.php?topic=1500.msg8110#msg8110), "We cannot estimate the SNR directly because the noise introduces uncertainty in the data," in 2009, and it maybe be so, I would not rely too much of SNR for weighting subframes.

Be very much interested in knowing if Juan's assessment is still valid!!! I use a combination of other subframe parameters, including #stars, star eccentricity, and FWHM.

[MineralMike]

I would not rely too much of SNR for weighting subframes.

As a PI newbie, I'm following instructions from Keller's book to use the sub with the highest SNR weight as the reference when integrating subs.

Is there a better choice? What should I use instead?

Thanks.

--- Mike

[pfile]

i uh... dont know what to say about that advice... the default method for subframe weighting in ImageIntegration is indeed SNR weight.

https://pixinsight.com/doc/tools/ImageIntegration/ImageIntegration.html#description_002

even juan's post linked above says that you can estimate SNR improvement between two images.

and per that post, as long as you can properly scale/normalize two images, you can indeed compare the SNR between the two images. and that's what ImageIntegration does in the absence of any externally generated weighting keywords.

rob

[MineralMike]

the default method for subframe weighting in ImageIntegration is indeed SNR weight.

Thanks. In the SubframeSelector, should I sort the list of subs in descending or ascending order on SNR Weight to choose which sub to use as the reference in imageintegration?

IOW, is a high SNR Weight better than a low SNR weight? I know little about statistics, so I'm searching for simple guidelines.

Thank you.

--- Mike

[pfile]

well i think larger SNRWeights are better but since i don't regularly use BPP or WBPP i am just going by what i think the documentation means. maybe someone else can say more definitively.

rob

[aworonow]

I agree that SNRWeights appear to be more in tune with what one would want from subframes than other noise measures.