A Word doc for Subframe Selector

[mschuster]

David,

Thanks for your results.

For four targets, I compared an SNRWeight integration to an SNRWeight / FWHM^2 integration, using the same set of accepted frames and the same reference frame for both.

For each target, the integrations are nearly identical (statistics, star measurements, and histograms). Visually they look identical.

Maybe this is because the FWHM variance in my accepted frames is small (I reject poor FWHM frames), and because small weight differences are averaged out in the integrations of many frames (20+), and because my setup is under-sampled which tends to minimize seeing FWHM variations naturally.

Probably my frames are not a good test case for these weighting comparisons.

Thanks,
Mike

PS: One advantage of these types of weightings is that they don't depend on frame set statistics. Frame weight is a function of the frame's data only, which is a nice, simplifying property.

[dnault42]

That's a good point Mike.  Data sets with small variations in FWHM would probably be better served by just setting the weight to SNRWeight.

Here's the updated spreadsheet:
https://drive.google.com/file/d/0B2J4InZni9UrTW13bWRUZEZla1U/view?usp=sharing

Regards,
David

[mschuster]

Just to double check:
- Weights must be positive. A zero or negative weight would be an error.
- Weights work in ratios. Given weighs w1 and w2 for two frames, the ratio w1 / w2 should typically be in the range 0.25 to 4. IMO it would be rare to have one frame weighted more than 4 times another. If more than 4 or so I would rather just reject the bad frame.

Mike

[dnault42]

I'd agree with that Mike.  That's one of the reasons I prefer equations where I can control the output range.  My default settings typically has a 2x range usually going from 50 to 100, making it a little easier to see.  That gets scaled (not normalized) by ImageIntegration when the files are read in.

Regards,
David