FWHM

[stefanocovino]

Hello everybody,

this is not a bug report. I am just asking wether there is a way to derive an estimate of the FWHM of the sources extracted by "sep" more or less as SExtractor does.

Thanks,
Stefano

[martinberoiz]

I would like to know this too. SEP returns a, b and theta, I guess one could choose max(a,b) as FWHM? (I'm not a contributor to sep)

[stefanocovino]

My bet would be something like:

FWHM = 2 * sqrt ( ln(2) * (x2 + y2))

where x2 and y2 are the second order moment (assuming Gaussian shape, etc. etc.).

But I would like to know what the sep developers think.

Stefano

---

Mobilis in mobile

Il giorno 08 mar 2016, alle ore 20:45, Martin Beroiz notifications@github.com ha scritto:

I would like to know this too. SEP returns a, b and theta, I guess one could choose max(a,b) as FWHM? (I'm not a contributor to sep)

—
Reply to this email directly or view it on GitHub.

[kbarbary]

sep doesn't have the equivalent of source extractor's FWHM_IMAGE / FWHM_WORLD parameter, though I think it would be a good addition.

Currently, you can use some approximation you suggest, but I'd probably do it in terms of a and b to take into account any correlation between x and y (the SourceExtractor manual describes a and b as "the maximum and minimum spatial RMS"). So, FWHM = 2 * sqrt(ln(2) * (a^2 + b^2))

I'll leave this open as a reminder to look into adding the SoureExtractor-style FWHM calculation to sep.

[cmccully]

The other fun thing is that for a Gaussian source (which may or may not be a good assumption) the 50% flux radius is the HWHM.

[kbarbary]

I worked on this a bit and only got it partially working, in that I got about half the sources in the test image to agree with FWHM values from Source Extractor. The other half are off by varying degrees (sometimes a lot!) and it's not clear why. The code is copied from Source Extractor. One difference is that SE treats saturated pixels, but SEP does not. This doesn't seem to be the root of the FWHM differences though.

I probably won't work on this any more, but my progress is in the 822f553 commit on the fwhm branch in this repo if anyone wants to take it over and dig deeper.

[rrgupta]

I think the issue might be related to the difference between the windowed and un-windowed 2nd moments. The FWHM should be the same for bright stars and faint stars, but I find that the un-windowed A and B parameters for stars increase with flux and so will not give reliable measures of FWHM. I think computing windowed 2nd moments analogous to SExtractor's X2WIN_IMAGE, Y2WIN_IMAGE, XYWIN_IMAGE (and the corresponding A and B from those) might do the trick.