search

lofar-astron / PyBDSF

Python Blob Detector and Source Finder
https://pybdsf.readthedocs.io
GNU General Public License v3.0
50 stars 24 forks source link

Peak flux greater than total flux? #42

Open cyriltasse opened 7 years ago

cyriltasse commented 7 years ago

Simulating an image with point source, I was very surprised to see that the estimated peak flux could be greater than the total flux... Is that an expected behaviour?

image

darafferty commented 7 years ago

Yeah, it is allowed to fit Gaussians smaller than the restoring beam, so this can happen (especially with low S/N sources). You can freeze the size of the Gaussians to the beam with fix_to_beam = True if you know all sources are point sources.

cyriltasse commented 7 years ago

But isn't pybdsf fitting for Gaussian(Total_flux_deconv,Sig_maj_deconv,Sig_min_deconv) x restoring_beam where x is a convolution? In that case how is it even possible? Or maybe it's fitting for Gaussian(Total_flux,Sig_maj,Sig_min) and then given restoring_beam it's computing Total_flux_deconv,Sig_maj_deconv,Sig_min_deconv,Peak_flux_deconv thresholding Sig_maj_deconv,Sig_min_deconv. Cos I don't get consistent values between Total_flux/Peak_flux and Sig_maj_deconv*Sig_min_deconv/(restoring_beam**2). I'm very confused...

darafferty commented 7 years ago

Yes, the second method is used: it fits the source then calculates the total flux as peak*size[0]*size[1]/(bm_pix[0]*bm_pix[1]) where bm_pix is the restoring beam size and size is given by Sig_maj and Sig_min (i.e., not the deconvolved size).

cyriltasse commented 7 years ago

ok - sorry - I made a mistake above - so I meant Sig_maj*Sig_min/(restoring_beam**2) and by Sig one should understand FWHM. I'll double check - but I wasn't finding total=peak*size[0]*size[1]/(bm_pix[0]*bm_pix[1])