def find_rotation(
imagedata,
rotdegs,
mx,
my,
sx,
sy,
xo,
yo, # for Affine2d
pixel,
npix,
bandpass,
over,
holeshape,
outdir=None): # for nrm_model
""" AS AZG 2018 08 Ann Arbor Develop the rotation loop first """
vprint("Before Loop: ", rotdegs)
#Extend this name to include multi-paramater searches?
psffmt = 'psf_nrm_{0:d}_{1:s}_{2:.3f}um_r{3:.3f}deg.fits'
# expect (npix, holeshape, bandpass/um, scl)
if hasattr(rotdegs, '__iter__') is False:
rotdegs = (rotdegs, )
affine2d_list = create_afflist_rot(rotdegs, mx, my, sx, sy, xo, yo)
crosscorr_rots = []
for (rot, aff) in zip(rotdegs, affine2d_list):
vprint(aff.name + "...")
jw = NRM_Model(mask='jwst',
holeshape=holeshape,
over=over,
affine2d=aff)
jw.set_pixelscale(pixel)
jw.simulate(fov=npix, bandpass=bandpass, over=over)
psffn = psffmt.format(npix, holeshape, bandpass / um, rot)
if outdir:
fits.PrimaryHDU(data=jw.psf).writeto(outdir + "/" + psffn,
overwrite=True)
fits.writeto(psffn, jw.psf, overwrite=True)
header = fits.getheader(psffn)
utils.affinepars2header(header, aff)
fits.update(psffn, jw.psf, header=header)
crosscorr_rots.append(utils.rcrosscorrelate(imagedata, jw.psf).max())
del jw
vprint("Debug: ", crosscorr_rots, rotdegs)
rot_measured_d, max_cor = utils.findpeak_1d(crosscorr_rots, rotdegs)
vprint("Rotation measured: max correlation {1:.3e}", rot_measured_d,
max_cor)
# return convenient affine2d
return utils.Affine2d(rotradccw=np.pi * rot_measured_d / 180.0,
name="{0:.4f}".format(rot_measured_d))
def find_scale(
imagedata,
affine_best, # best current guess at data geometry cf analytical ideal
scales, # scales are near-unity
pixel,
npix,
bandpass,
over,
holeshape,
outdir=None): # for nrm_model
""" Preserve incoming "pixel" value, put the scale correction into the Affine2d object
Is that kosher??? Should we change the pixel scale and leave affine2d the same?
Affine2d can also incorporate unequal x and y scales, shears...
For now place scale corrections into the Affine2d object
Note - placing isotropic scale change into Affine2d is equivalent to changing
the effective image distance in the optical train while insisting that the
mask physical geometry does not change, and the wavelength is perfectly knowm
AS 2018 10 """
affine_best.show("\tfind_scale")
vprint("\tBefore Loop: ", scales)
#Extend this name to include multi-paramater searches?
psffmt = 'psf_nrm_{0:d}_{1:s}_{2:.3f}um_scl{3:.3f}.fits'
# expect (npix, holeshape, bandpass/um, scl)
if hasattr(scales, '__iter__') is False:
scales = (scales, )
affine2d_list = create_afflist_scales(scales, affine_best.mx,
affine_best.my, affine_best.sx,
affine_best.sy, affine_best.xo,
affine_best.yo)
crosscorrs = []
for (scl, aff) in zip(scales, affine2d_list):
vprint(aff.name + "...")
jw = NRM_Model(mask='jwst',
holeshape=holeshape,
over=over,
affine2d=aff)
jw.set_pixelscale(pixel)
jw.simulate(fov=npix, bandpass=bandpass, over=over)
psffn = psffmt.format(npix, holeshape, bandpass[:, 1][0] / um, scl)
if outdir:
fits.PrimaryHDU(data=jw.psf).writeto(outdir + "/" + psffn,
overwrite=True)
fits.writeto(psffn, jw.psf, overwrite=True)
header = fits.getheader(psffn)
utils.affinepars2header(header, aff)
fits.update(psffn, jw.psf, header=header)
crosscorrs.append(utils.rcrosscorrelate(imagedata, jw.psf).max())
del jw
vprint("\tfind_affine2d_parameters: crosscorrelations", crosscorrs)
vprint("\tfind_affine2d_parameters: scales", scales)
scl_measured, max_cor = utils.findpeak_1d(crosscorrs, scales)
vprint(
"\tfind_affine2d_parameters factor measured {0:.5f} Max correlation {1:.3e}"
.format(scl_measured, max_cor))
vprint("\tfind_affine2d_parameters pitch from header {0:.3f} mas".format(
pixel * rad2mas))
vprint(
"\tfind_affine2d_parameters pitch {0:.3f} mas (implemented using affine2d)"
.format(scl_measured * pixel * rad2mas))
# return convenient affine2d
return utils.Affine2d(affine_best.mx * scl_measured,
affine_best.my * scl_measured,
affine_best.sx * scl_measured,
affine_best.sy * scl_measured,
affine_best.xo * scl_measured,
affine_best.yo * scl_measured,
name="scale_{0:.4f}".format(scl))
def run_data_correlate(data, model):
sci = data
print("shape sci", np.shape(sci))
print("shape model", np.shape(model))
return utils.rcrosscorrelate(sci, model)