def spline_coeff_eval(raw_img,hcenters,hc_ref,vcenters,vc_ref,invar,r_breakpoints,spline_poly_coeffs,s_scale,sigmas,powers,cpad=5,bp_space=2,full_image=True,view_plot=False,sp_coeffs=None,ecc_pa_coeffs=None):
""" Another highly specialized function. Evaluates coeffs found in
sf.interpolate_coeffs. Right now, displays images of the fit
compared to data.
Returns spline_fit
"""
voff = 1
for k in range(len(vcenters)):
if full_image:
hmean = hcenters[k]
vmean = vcenters[k]
small_img = raw_img
small_inv = invar
else:
harr = np.arange(-cpad,cpad+1)+hcenters[k]
varr = np.arange(-cpad,cpad+1)+vcenters[k]
small_img = raw_img[varr[0]:varr[-1]+1,harr[0]:harr[-1]+1]
small_inv = invar[varr[0]:varr[-1]+1,harr[0]:harr[-1]+1]
hmean, hheight, hbg = sf.fit_mn_hght_bg(harr,small_img[cpad,:],small_inv[cpad,:],sigmas[k],hcenters[k],sigmas[k],powj=powers[k])
vmean, vheight, vbg = sf.fit_mn_hght_bg(varr+voff,small_img[:,cpad],small_inv[:,cpad],sigmas[k],vcenters[k],sigmas[k],powj=powers[k])
hdec, hint = math.modf(hmean)
vdec, vint = math.modf(vmean)
# small_img = recenter_img(small_img,[vmean,hmean],[varr[0]+cpad,harr[0]+cpad])
# print "Mean is [", vmean, hmean, "]"
# plt.imshow(small_img,extent=(-cpad+hcenters[k],cpad+hcenters[k]+1,-cpad+vcenters[k],cpad+vcenters[k]+1),interpolation='none')
# plt.show()
# plt.close()
# r_breakpoints = [0, 1, 2, 3, 4, 5, 9]
# r_breakpoints = [0, 1.2, 2.5, 3.5, 5, 9]
# theta_orders=[0,-2,2]
theta_orders = [0]
# spline_fit = spline.spline_2D_radial(small_img,small_inv,)
# v_bpts = varr[np.mod(np.arange(len(varr)),bp_space)==0]-vcenters[k]
# h_bpts = harr[np.mod(np.arange(len(harr)),bp_space)==0]-hcenters[k]
spline_coeffs = np.zeros((len(spline_poly_coeffs[0])))
for l in range(len(spline_poly_coeffs[0])):
spline_coeffs[l] = sf.eval_polynomial_coeffs(hcenters[k],spline_poly_coeffs[:,l])
ecc_pa = np.ones((2))
if ecc_pa_coeffs is not None:
for m in range(2):
ecc_pa[m] = sf.eval_polynomial_coeffs(hcenters[k],ecc_pa_coeffs[m])
# if k==0:
# print spline_coeffs
params = lmfit.Parameters()
params.add('vc', value = vmean-vc_ref)
params.add('hc', value = hmean-hc_ref)
params.add('q',ecc_pa[0])
params.add('PA',ecc_pa[1])
spline_fit = spline.spline_2D_radial(small_img,small_inv,r_breakpoints,params,theta_orders=theta_orders,spline_coeffs=spline_coeffs,sscale=s_scale[k])
if view_plot:
plt.close()
res = (small_img-spline_fit)*np.sqrt(small_inv)
print("Chi^2 reduced = {}".format(np.sum(res**2)/(np.size(res)-2-len(spline_poly_coeffs[0]))))
vis = np.vstack((small_img,spline_fit,res))
plt.imshow(vis,interpolation='none')
# plt.imshow(spline_fit,interpolation='none')
plt.show()
plt.close()
plt.plot(spline_fit[:,5])
plt.show()
plt.close()
## plt.figure()
## plt.hist(np.ravel((small_img-spline_fit)*np.sqrt(small_inv)))
## plt.imshow((small_img-spline_fit)*small_inv,interpolation='none')
# chi2 = np.sum((small_img-spline_fit)**2*small_inv)/(np.size(small_img)-len(spline_coeffs))
# print("Chi^2 = {}".format(chi2))
# plt.show()
# plt.close()
return spline_fit
def spline_coeff_fit(raw_img,hcenters,vcenters,invar,r_breakpoints,sigmas,powers,theta_orders=[0],cpad=5,bp_space=2,return_new_centers=False):
""" Highly specialized function. Might want to re-write later for
generality. Takes an arc image and the horizontal/vertical centers
of known "good" peaks with known std (sigmas) and gauss power (powers)
Fits a spline to each peak and returns and array with the spline
coefficients (which can later be used for 2D extraction)
"""
new_hcenters = np.zeros(len(hcenters))
new_vcenters = np.zeros(len(vcenters))
fit_params = np.zeros((2,len(vcenters)))
scale_fit = zeros((len(vcenters)))
voff = 1
for k in range(len(vcenters)):
harr = np.arange(-cpad,cpad+1)+hcenters[k]
varr = np.arange(-cpad,cpad+1)+int(np.floor(vcenters[k]))
harr = harr[harr>=0]
harr = harr[harr<raw_img.shape[1]]
varr = varr[varr>=0]
varr = varr[varr<raw_img.shape[0]]
small_img = raw_img[varr[0]:varr[-1]+1,harr[0]:harr[-1]+1]
small_inv = invar[varr[0]:varr[-1]+1,harr[0]:harr[-1]+1]
hcut = int(np.mod(np.argmax(small_img),small_img.shape[1]))
vcut = int(np.floor(np.argmax(small_img)/small_img.shape[1]))
hmean, hheight, hbg = sf.fit_mn_hght_bg(harr,small_img[vcut,:],small_inv[vcut,:],sigmas[k],hcenters[k],sigmas[k],powj=powers[k])
vmean, vheight, vbg = sf.fit_mn_hght_bg(varr+voff,small_img[:,hcut],small_inv[:,hcut],sigmas[k],vcenters[k],sigmas[k],powj=powers[k])
hdec, hint = math.modf(hmean)
vdec, vint = math.modf(vmean)
# small_img = recenter_img(small_img,[vmean,hmean],[varr[0]+cpad,harr[0]+cpad])
# print "Mean is [", vmean, hmean, "]"
# plt.imshow(small_img,extent=(-cpad+hcenters[k],cpad+hcenters[k]+1,-cpad+vcenters[k],cpad+vcenters[k]+1),interpolation='none')
# plt.show()
# plt.close()
# r_breakpoints = [0, 1, 2, 2.5, 3, 3.5, 4, 5, 10]
# r_breakpoints = [0, 1, 2, 3, 4, 5, 9]
# theta_orders=[0,-2,2]
args = (small_img,small_inv,r_breakpoints)
kws = dict()
kws['theta_orders'] = theta_orders
params = lmfit.Parameters()
params.add('vc', value = vmean-varr[0])
params.add('hc', value = hmean-harr[0])
params.add('q',value = 0.85, min=0)
params.add('PA',value = 0)
minimizer_results = lmfit.minimize(spline.spline_residuals,params,args=args,kws=kws)
ecc = minimizer_results.params['q'].value
pos_ang = minimizer_results.params['PA'].value
if ecc > 1:
ecc = 1/ecc
pos_ang -= (np.pi/2)
pos_ang = pos_ang % (2*np.pi)
# print "Eccentricity = ", ecc
# print "Position Angle = ", pos_ang*180/(np.pi)
# center=[vmean-varr[0],hmean-harr[0]]
# print center
spline_fit, s_coeffs, s_scale = spline.spline_2D_radial(small_img,small_inv,r_breakpoints,minimizer_results.params,theta_orders=theta_orders,return_coeffs=True)
# v_bpts = varr[np.mod(np.arange(len(varr)),bp_space)==0]-vcenters[k]
# h_bpts = harr[np.mod(np.arange(len(harr)),bp_space)==0]-hcenters[k]
# spline_fit, s_coeffs, s_scale = spline.spline_2D(small_img,1/(small_img+readnoise**2),h_bpts,v_bpts,return_coeffs=True)
# if k == 0:
# vis = np.hstack((small_img,spline_fit,small_img-spline_fit))
# plt.imshow(vis,interpolation='none')
## # plt.figure()
## # plt.hist(np.ravel((small_img-spline_fit)*small_inv))
# plt.show()
# plt.close()
if k==0:
spline_coeffs = zeros((len(vcenters),len(s_coeffs)))
spline_coeffs[k] = s_coeffs
scale_fit[k] = s_scale
new_hcenters[k] = hmean
new_vcenters[k] = vmean
fit_params[:,k] = np.array(([ecc,pos_ang]))
if return_new_centers:
return spline_coeffs, scale_fit, fit_params, new_hcenters, new_vcenters
else:
return spline_coeffs, scale_fit, fit_params
