def run_pa(self,
input_frame,
psfbootfile,
outfile,
usesigma,
linelist=None,
npoly=2,
nbins=2,
domain=None):
from desispec.quicklook.arcprocess import process_arc, write_psffile
from desispec.quicklook.palib import get_resolution
from desispec.psf import PSF
wcoeffs = process_arc(input_frame,
linelist=linelist,
npoly=npoly,
nbins=nbins,
domain=domain)
#- write out the psf outfile
wstep = input_frame.meta["WAVESTEP"]
write_psffile(psfbootfile, wcoeffs, outfile, wavestepsize=wstep)
log.debug("Wrote psf file {}".format(outfile))
#- update the arc frame resolution from new coeffs
newpsf = PSF(outfile)
input_frame.resolution_data = get_resolution(input_frame.wave,
input_frame.nspec,
newpsf,
usesigma=usesigma)
return input_frame
def do_boxcar(image,
psf,
outwave,
boxwidth=2.5,
nspec=500,
maskFile=None,
usesigma=False):
"""Extracts spectra row by row, given the centroids
Args:
image : desispec.image object
psf: desispec.psf.PSF like object
Or do we just parse the traces here and write a separate wrapper to handle this? Leaving psf in the input argument now.
outwave: wavelength array for the final spectra output
boxwidth: HW box size in pixels
usesigma: if True, use sigma from psfboot file(xsigma) or psf file (wsigma) to calculate resolution data.
Returns flux, ivar, resolution
"""
import math
from desispec.frame import Frame
#wavelength=psf.wavelength() # (nspec,npix_y)
def calcMask(psf):
wmin = psf.wmin
wmax = psf.wmax
waves = np.arange(wmin, wmax, 0.25)
xs = psf.x(None, waves) #- xtraces # doing the full image here.
ys = psf.y(None, waves) #- ytraces
camera = image.camera
spectrograph = int(camera[1:]) #- first char is "r", "b", or "z"
imshape = image.pix.shape
mask = np.zeros((imshape[1], imshape[0]))
maxx, maxy = mask.shape
maxx = maxx - 1
maxy = maxy - 1
ranges = np.zeros((mask.shape[1], xs.shape[0] + 1), dtype=int)
for bin in range(0, len(waves)):
ixmaxold = 0
for spec in range(0, xs.shape[0]):
xpos = xs[spec][bin]
ypos = int(ys[spec][bin])
if xpos < 0 or xpos > maxx or ypos < 0 or ypos > maxy:
continue
xmin = xpos - boxwidth
xmax = xpos + boxwidth
ixmin = int(math.floor(xmin))
ixmax = int(math.floor(xmax))
if ixmin <= ixmaxold:
print("Error Box width overlaps,", xpos, ypos, ixmin,
ixmaxold)
return None, None
ixmaxold = ixmax
if mask[int(xpos)][ypos] > 0:
continue
# boxing in x vals
if ixmin < 0: #int value is less than 0
ixmin = 0
rxmin = 1.0
else: # take part of the bin depending on real xmin
rxmin = 1.0 - xmin + ixmin
if ixmax > maxx: # xmax is bigger than the image
ixmax = maxx
rxmax = 1.0
else: # take the part of the bin depending on real xmax
rxmax = xmax - ixmax
ranges[ypos][spec + 1] = math.ceil(xmax) #end at next column
if ranges[ypos][spec] == 0:
ranges[ypos][spec] = ixmin
mask[ixmin][ypos] = rxmin
for x in range(ixmin + 1, ixmax):
mask[x][ypos] = 1.0
mask[ixmax][ypos] = rxmax
for ypos in range(ranges.shape[0]):
lastval = ranges[ypos][0]
for sp in range(1, ranges.shape[1]):
if ranges[ypos][sp] == 0:
ranges[ypos][sp] = lastval
lastval = ranges[ypos][sp]
return mask, ranges
if maskFile is not None:
import os
if os.path.exists(maskFile) and os.path.isfile(maskFile):
f = open(maskFile, 'rb')
npf = np.load(f)
mask = npf['mask']
ranges = npf['ranges']
print("Loading mask from file %s" % maskFile)
else:
print(
"Mask file is given but doesn't exist. Generating mask and saving to file %s"
% maskFile)
mask, ranges = calcMask(psf)
try:
f = open(maskFile, 'wb')
np.savez(f, mask=mask, ranges=ranges)
except:
pass
else:
mask, ranges = calcMask(psf)
maskedimg = (image.pix * mask.T)
maskedvar = (1 / image.ivar.clip(0) * mask.T)
flux = np.zeros((maskedimg.shape[0], ranges.shape[1] - 1))
ivar = np.zeros((maskedimg.shape[0], ranges.shape[1] - 1))
for r in range(flux.shape[0]):
row = np.add.reduceat(maskedimg[r], ranges[r])[:-1]
flux[r] = row
vrow = np.add.reduceat(maskedvar[r], ranges[r])[:-1]
ivar[r] = 1 / vrow
wtarget = outwave
#- limit nspec to psf.nspec max
if nspec > psf.nspec:
nspec = psf.nspec
print("Warning! Extracting only {} spectra".format(psf.nspec))
fflux = np.zeros((nspec, len(wtarget)))
iivar = np.zeros((nspec, len(wtarget)))
#- convert to per angstrom first and then resample to desired wave length grid.
for spec in range(nspec):
ww = psf.wavelength(spec)
dwave = np.gradient(ww)
flux[:, spec] /= dwave
ivar[:, spec] *= dwave**2
fflux[spec, :], iivar[spec, :] = resample_spec(ww, flux[:, spec],
wtarget, ivar[:, spec])
#- Get resolution from the psf
resolution = get_resolution(wtarget, nspec, psf, usesigma=usesigma)
return fflux, iivar, resolution
def do_boxcar(image,tset,outwave,boxwidth=2.5,nspec=500,maskFile=None,usesigma=False,
quick_resolution=False):
"""Extracts spectra row by row, given the centroids
Args:
image : desispec.image object
tset: desispec.xytraceset like object
outwave: wavelength array for the final spectra output
boxwidth: HW box size in pixels
usesigma: if True, use sigma from psf file (ysigma) to calculate resolution data.
quick_resolution: whether to calculate the resolution matrix or use QuickResolution object
Returns flux, ivar, resolution
"""
import math
from desispec.frame import Frame
#wavelength=psf.wavelength() # (nspec,npix_y)
def calcMask(tset):
wmin=tset.wavemin
wmax=tset.wavemax
waves=np.arange(wmin,wmax,0.25)
xs=tset.x_vs_wave(np.arange(tset.nspec),waves) #- xtraces # doing the full image here.
ys=tset.y_vs_wave(np.arange(tset.nspec),waves) #- ytraces
camera=image.camera
spectrograph=int(camera[1:]) #- first char is "r", "b", or "z"
imshape=image.pix.shape
mask=np.zeros((imshape[1],imshape[0]))
maxx,maxy=mask.shape
maxx=maxx-1
maxy=maxy-1
ranges=np.zeros((mask.shape[1],xs.shape[0]+1),dtype=int)
for bin in range(0,len(waves)):
ixmaxold=0
for spec in range(0,xs.shape[0]):
xpos=xs[spec][bin]
ypos=int(ys[spec][bin])
if xpos<0 or xpos>maxx or ypos<0 or ypos>maxy :
continue
xmin=xpos-boxwidth
xmax=xpos+boxwidth
ixmin=int(math.floor(xmin))
ixmax=int(math.floor(xmax))
if ixmin <= ixmaxold:
print("Error Box width overlaps,",xpos,ypos,ixmin,ixmaxold)
return None,None
ixmaxold=ixmax
if mask[int(xpos)][ypos]>0 :
continue
# boxing in x vals
if ixmin < 0: #int value is less than 0
ixmin=0
rxmin=1.0
else:# take part of the bin depending on real xmin
rxmin=1.0-xmin+ixmin
if ixmax>maxx:# xmax is bigger than the image
ixmax=maxx
rxmax=1.0
else: # take the part of the bin depending on real xmax
rxmax=xmax-ixmax
ranges[ypos][spec+1]=math.ceil(xmax)#end at next column
if ranges[ypos][spec]==0:
ranges[ypos][spec]=ixmin
mask[ixmin][ypos]=rxmin
for x in range(ixmin+1,ixmax): mask[x][ypos]=1.0
mask[ixmax][ypos]=rxmax
for ypos in range(ranges.shape[0]):
lastval=ranges[ypos][0]
for sp in range(1,ranges.shape[1]):
if ranges[ypos][sp]==0:
ranges[ypos][sp]=lastval
lastval=ranges[ypos][sp]
return mask,ranges
if maskFile is not None:
import os
if os.path.exists(maskFile) and os.path.isfile(maskFile):
f=open(maskFile,'rb')
npf=np.load(f)
mask=npf['mask']
ranges=npf['ranges']
print("Loading mask from file %s"%maskFile)
else:
print("Mask file is given but doesn't exist. Generating mask and saving to file %s"%maskFile)
mask,ranges=calcMask(tset)
try:
f=open(maskFile,'wb')
np.savez(f,mask=mask,ranges=ranges)
except:
pass
else:
mask,ranges=calcMask(tset)
Tmask=mask.T
maskedimg=(image.pix*Tmask)
maskedvar=(Tmask/image.ivar.clip(1e-8))
flux=np.zeros((maskedimg.shape[0],ranges.shape[1]-1))
ivar=np.zeros((maskedimg.shape[0],ranges.shape[1]-1))
for r in range(flux.shape[0]):
row=np.add.reduceat(maskedimg[r],ranges[r])[:-1]
flux[r]=row
vrow=np.add.reduceat(maskedvar[r],ranges[r])[:-1]
ivar[r]=1/vrow
wtarget=outwave
#- limit nspec to tset.nspec max
if nspec > tset.nspec:
nspec=tset.nspec
print("Warning! Extracting only {} spectra".format(tset.nspec))
fflux=np.zeros((nspec,len(wtarget)))
iivar=np.zeros((nspec,len(wtarget)))
#- convert to per angstrom first and then resample to desired wave length grid.
for spec in range(nspec):
ww=tset.wave_vs_y(spec,np.arange(0,tset.npix_y))
dwave=np.gradient(ww)
flux[:,spec]/=dwave
ivar[:,spec]*=dwave**2
fflux[spec,:],iivar[spec,:]=resample_spec(ww,flux[:,spec],wtarget,ivar[:,spec])
#- Get resolution from the psf
resolution=get_resolution(wtarget,nspec,tset,usesigma=usesigma)
return fflux,iivar,resolution