Open lbaumo opened 1 year ago
add some modification of these functions to the vignet module:
` def get_sigma_size(cat):
fwhm_fac = 2*numpy.sqrt(2*numpy.log(2))
ellipticity = 1.0 - cat['b_world']/cat['a_world']
sigma = cat['flux_radius']*2/fwhm_fac
drad = sigma*5
drad = drad*(1.0 + ellipticity)
drad = numpy.ceil(drad)
sigma_size = 2*drad.astype('i4') # sigma size is twice the radius
print "returning sigma size"
return sigma_size
`
`def get_box_sizes(cat):
#get box sizes that are wither 2**N or 3*2**N, within
#the limits set by the user
min_box_size = 32
max_box_size = 256
sigma_size = get_sigma_size(cat)
# now do row and col sizes
row_size = cat['ymax_image'] - cat['ymin_image'] + 1
col_size = cat['xmax_image'] - cat['xmin_image'] + 1
# get max of all three
box_size = numpy.vstack((col_size,row_size,sigma_size)).max(axis=0)
# clip to range
box_size = box_size.clip(min_box_size,max_box_size)
return box_size
`
That is a good idea!
Just a small warning, all the code has been done with odd size postage stamp. As you have written right now the sigma_size
will always be an even number. It is possible that it creates some issues latter on. That needs to be checked.
Are we losing bright galaxies because the postage stamp is too small? We should have the postage stamp scale with the galaxy size above a certain threshold.