FWHM horizontal to vertical ratio filtering

[lowenthalm]

The FWHM filter in live stack is great for catching windy frames, bad tracking frames and bad star bloat frames when your neighbor's furnace plume turns on. However, its a bit of a blunt instrument. Variations in seeing are normal, with the effects of bad tracking and windy frames superimposed on this natural variation. I have to keep an eye on FWHM because I need to keep it just a bit above average seeing variation in order to reject windy and bad track subs. I usually observe what the minimum FWHM is currently and pick a FWHM filter value roughly 10% higher than this observed FWHM minimum (my FWHM is typically 3 to 7 at B1x1 and I switch to B2x2 and image larger and less detailed class deep sky objects if its above 7 or 8). Values higher than about 10% of the recent FWHM minima are usually from oblong stars caused by wind and tracking errors.

It would be nice if the FWHM filter could be modified so that these two classes of problems are filtered using different controls in the FWHM filter panel. One would filter on the average FWHM as it does now. I suggest adding a second pair of FWHM filter control values to filter based on the ratio of FWHM width divided by FWHM height. With this feature, I could then filter wind and tracking errors by rejecting frames with oblong stars with a width to height ratio outside a range, say, 0.77 to 1.3. I could then set the average FWHM filter at a much higher value that would filter frames when seeing really gets horrible. This would require a lot less supervision, chasing seeing up and down a pixel or two to keep filtering out wind and bad tracking and keeping only frames with round stars.

[admin]

Hi,

I do actually have a roundness filter already on my to-do list, which will be separate to the current FWHM. This will I think help catch those frames that are either wind affected or have mount drift, etc. Hopefully that will cover a lot of the ground for you.

cheers,

Robin

[lowenthalm]

Yep, a roundness filter would do the trick. This will be awesome for imaging with a big tracking dob too! I used one of these for a few years before I finally bought myself a big SCT late last year. I was just trying to think within the confines of the current implementation. I figured both the X and Y FWHM must be in there somewhere just itching to get filtered with some math.

[admin]

Hi,

the code I have already filters to some extend by circularity, so stars will tend not to be detected if they are too extended (which will then tend to drop the frame due to a failure to align if enough stars drop out for this reason). I already post-process each detected star to re-calculate the FWHM and to look for things like multiple nearby peaks in brightness which would indicate a binary/group and therefore invalidate the FWHM calculation. I should be able to add some sort of circularity measurement to that.

It will be interesting to see how well it works - I imagine that a fully smeared frame (stars are lines) will be fairly easy to cope with, but what about a frame where there is bad movement during 10% of the time? 90% of the brightness will be properly concentrated in a gaussian star - will the remaining 10% be enough to triggger a significant change in circularity?

cheers,

Robin