Integration using FWHM weights?

[MortenBalling]

Hi All

Merry Xmas!

I've been experimenting with something I call "Crowd Imaging". The idea is to take collaborations to the max. I try to do so, by integrating several hundreds of Creative Commons licensed astro photos into one stack with very high SNR. All that works fine.

Lately I've been experimenting with the SubFrameSelector script, sorting out half of the images with the lowest FWHM, and then stacking those separately. That also works perfectly.

However, I've seen the possibility to add fits keywords using SFS. Instead of sorting, I'd like to add FWHM to each file using SFS, and then integrate all files with FWHM weight.

Is that possible, and how do I do that?

I know how to select a fits keyword for weights in ImageIntegration, but by using FWHM, won't PI weigh the files with the highest FWHM higher than the ones with low FWHM?

Thanks in advance.

Best regards

Morten

Ps. If you wan't to see some examples of the Crowd Image method you can see more here:

http://www.astrobin.com/users/MortenBalling/

[MortenBalling]

I guess I found a way, but I'll probably have to improve my formula

After measuring all subframes in SFS, I made a very simple weight expression:

WEIGHT = 1/FWHM

It seems to work when I then integrate using fits keyword = WEIGHT, but the formula is a bit rough.

[MortenBalling]

A picture tells more than a thousand words, so here are the same 30 sub images integrated with/without FWHM weights:

[georg.viehoever]

A picture tells more than a thousand words, so here are the same 30 sub images integrated with/without FWHM weights:
...

Using weights during the integration usually has the goal to minimize noise. Theory tells us that we get the best result when weighting with 1/(sigma^2).

I am not sure what the benefit of weights is in the context of FWHM. Of course, the integration is "sharper" if you give pictures with large FWHM a low weight. But you get the same result (or even better) by setting weight to 0 for those pictures that dont meet your FWHM standards. What do you gain by giving bad images a low weight instead of just excluding them from the stack?

Georg

[dayers]

Georg, how might WEIGHT = 1/FWHM compare with just using SNRWeight?

Dave

[MortenBalling]

@Georg

In the first post, I wrote that I'm working with a special technique, where I search the Internet for as many Creative Commons astro photos of one deep sky object I can find, and then combine all the images into one. The idea is to try and take collaborations even further.

Combining several hundred images, even though they are nonlinear JPGs, results in an amazing SNR due to the combined integration time. Often more than 500 hours in total.

Someone at the danish astronomy forum, mentioned that the method should be called Crowd Imaging, and that name stuck. I've done quite a few so far, but because it's sort of Big Data projects, it's pretty time consuming. I constantly find newer, smarter and faster ways to make it work.

Basically I gather as many images as I can find. Then I crop the edges, and flip the ones who need it. After that I StarAlign all the images to a single image with a large field.

Next I use SubFrameSelector to measure FWHM of the aligned frames. The images I use, vary quite a lot, both regarding color, sharpness, resolution, SNR etc. A funny thing though, is that people are pretty precise with how they rotate the field. I therefore normally end up with eight spikes.

The SFS FWHM weights are then used for ImageIntegration. In theory I should loose SNR, but in practice the images I find with low FWHM often have better SNR. I integrate using black clipping, add/scale normalization and no rejection.

You can see some results here:

http://www.astrobin.com/users/MortenBalling/

I wanted to share the FWHM weight idea, because it might come in handy in some other cases as well.

One is Lucky Imaging. I tried to shoot a 300x1sec frame series of Trapezuim with small pixel resolution, and sorted out the 200 with best eccentricity and FWHM. Then I integrated using weight=1/FWHM^2, and used Pixelmath to "replace" with the sharp stack on a normal stack, using the normal stack as mask. I need way more frames to make it work, but so far it looks promising. I can even see the proplyds.

Another way to use this is when you combine several nights of exposure, with different seeing, or when you use different exposure times. Then you mix in the sharp stack in the bright areas of a stack with max SNR, like mentioned above.

My formula is very primitive and homemade, so there's a lot of room for improvement. But I guess the first pancakes weren't perfect either

Best regards

Morten