def singlePartAnalysis(surf, img, pxlen, thres, h_tip, ar_tip, N_part, N_pxl,
param_string_list, binwidth_list, bin_min_list,
bin_max_list):
# get single particles
surf_obj_list, surf_labeled = mf.identObj(surf, thres)
img_obj_list, img_labeled = mf.identObj(img, thres)
mf.plotThres(
surf, surf_labeled, pxlen,
'surface (found ' + str(len(surf_obj_list)) + ' of ' + str(N_part) +
' particles, thres=' + str(thres) + 'nm)')
mf.plotThres(
img, img_labeled, pxlen, 'image (found ' + str(len(img_obj_list)) +
' of ' + str(N_part) + ' particles, thres=' + str(thres) + 'nm)')
surf_param_list = [
calcParams(obj_i, pxlen, thres) for obj_i in surf_obj_list
]
img_param_list = [
calcParams(obj_i, pxlen, thres) for obj_i in img_obj_list
]
# plot histograms of parameters
N_param = len(surf_param_list[0])
for i in range(N_param): # iterate over all parameters (mean, std, etc.)
surf_dist = [
surf_param_list[j][i] for j in range(len(surf_param_list))
]
img_dist = [img_param_list[j][i] for j in range(len(img_param_list))]
fig, (ax1, ax2) = plt.subplots(nrows=2,
ncols=1,
sharex=True,
sharey=True)
bins = np.arange(min(surf_dist + img_dist),
max(surf_dist + img_dist) + binwidth_list[i],
binwidth_list[i])
ax1.hist(surf_dist,
color='b',
bins=bins,
edgecolor='black',
linewidth=2)
ax2.hist(img_dist,
color='r',
bins=bins,
edgecolor='black',
linewidth=2)
ax1.set_xlim(bin_min_list[i], bin_max_list[i])
ax2.set_xlim(bin_min_list[i], bin_max_list[i])
ax2.set_xlabel(param_string_list[i])
ax1.set_ylabel('frequency')
ax2.set_ylabel('frequency')
ax1.set_title('surface (found ' + str(len(surf_obj_list)) + ' of ' +
str(N_part) + ' particles, thres=' + str(thres) + 'nm)')
ax2.set_title('image (found ' + str(len(img_obj_list)) + ' of ' +
str(N_part) + ' particles, thres=' + str(thres) + 'nm)')
fig.suptitle(r'$h_{tip}=$' + str(h_tip) + r'$nm,\ a.r.=$' +
str(ar_tip) + r'$,\ N_{pxl}=$' + str(N_pxl) +
r'$,\ l_{px}=$' + str(pxlen) + r'$nm,\ thres=$' +
str(thres) + r'$nm$')
def reconstLogNorm(z, pxlen, thres, N_part, R_mu_real, R_sigma_real):
z_obj_list, z_labeled = mf.identObj(z, thres)
mf.plotThres(z, z_labeled, pxlen, 'found ' + str(len(z_obj_list)) + ' of ' + str(N_part) + ' particles, thres=' + str(thres)+'nm')
R_list = [(np.max(np.shape(obj_i)))*pxlen/2 for obj_i in z_obj_list]
R_mean = np.mean(R_list)
R_std = np.std(R_list)
R_mu = np.log(R_mean / np.sqrt(1 + R_std**2 / R_mean**2)) # recalculated gaussian
R_sigma = np.sqrt(np.log(1 + (R_std/R_mean)**2)) # reculaculated gaussian
x = np.linspace(R_mean - 3*R_std, R_mean + 3*R_std, 1000)
pdf = 1 / (x * R_sigma * np.sqrt(2*np.pi)) * np.exp(-(np.log(x) - R_mu)**2 / (2*R_sigma**2))
pdf_real = 1 / (x * R_sigma_real * np.sqrt(2*np.pi)) * np.exp(-(np.log(x) - R_mu_real)**2 / (2*R_sigma_real**2))
plt.figure()
plt.hist(R_list, bins=12, density=True, edgecolor='black', linewidth=2, color='grey', alpha=0.5)
plt.plot(x, pdf, color='r', linewidth=3.5, label='empiric distribution (R_mu = {0:.3f}, R_sigma = {1:.3f})'.format(R_mu, R_sigma))
plt.plot(x, pdf_real, color='green', linewidth=3.5, label='real distribution (R_mu = {0:.3f}, R_sigma = {1:.3f})'.format(R_mu_real, R_sigma_real))
plt.xlabel(r'$R_{part} [nm]$')
plt.ylabel('frequency')
plt.title('found ' + str(len(z_obj_list)) + ' of ' + str(N_part) + ' particles, thres=' + str(thres)+'nm')
plt.legend(loc=1)
plt.tight_layout()
return R_mean, R_std, R_mu, R_sigma
#rtip=s*500
htip = hspikemax * 1.02
spikedist = 2 * np.sqrt(rtip**2 - (rtip - hspikemax)**2) + 2
#MAP-------------------------------------------------
# z=mf.genFlat(Npx)
# #z=mf.genNormNoise(z,pxlen, 2, 2)
# z=mf.genHexSpikes(z,pxlen,hspikemin,hspikemax,spikedist,rspikemin,rspikemax,
# 0.9,0.2,par='r', xmin=spikedist/2, xmax=len(z)*pxlen-spikedist/2,
# ymin=spikedist/2, ymax=len(z)*pxlen-spikedist/2)
# mf.plotfalsecol(z,pxlen)
# np.savetxt('map1.dat',z, header=str(pxlen)+' '+str(Npx)+' '+str(rtip)+' '+str(s)+' \n pxlen, Npx, Rtip, s')
# #Tip------------------------------------------------
# #occhio che h/a>>pxlen
# if hspikemax>rtip: print('Warning: possible spikes higher than semisphere tip')
# tip=mf.genSemisphTip(pxlen,htip,r=rtip)
# np.savetxt('tip1.dat',tip, header=str(pxlen)+' '+str(Npx)+' '+str(rtip)+' '+str(s)+' \n pxlen, Npx, Rtip, s')
# mf.plotfalsecol(tip,pxlen)
# # #IMG------------------------------------------------
# img = mph.grey_dilation(z, structure=-tip)
# img = mf.genNormNoise(img,pxlen, 4, 2)
# np.savetxt('img1.dat',img, header=str(pxlen)+' '+str(Npx)+' '+str(rtip)+' '+str(s)+' \n pxlen, Npx, Rtip, s')
# mf.plotfalsecol(img,pxlen)
# #---------------------------------------------------
img = np.loadtxt('img1.dat')
img = mph.filters.median_filter(img, 5)
img_obj_list, img_labeled, img_obj_ind = mf.identObj(img, thres, Npx_min=5)
mf.plotThres(img, img_labeled, pxlen)
def reconstLogNorm(z, pxlen, R_tip, sigma_noise, thres, N_part, R_median_real,
R_mu_real, R_sigma_real):
# z_obj_list, z_labeled, z_obj_ind = mf.identObj(z, thres)
# mf.plotThres(z, z_labeled, pxlen, 'found ' + str(len(z_obj_list)) + ' of ' + str(N_part) + ' particles, thres=' + str(thres)+'nm')
z_obj_list, z_labeled = mf.identObj_Laplace(z, thres)
mf.plotThres(
z, z_labeled, pxlen, 'found ' + str(len(z_obj_list)) + ' of ' +
str(N_part) + ' particles, thres=' + str(thres) + 'nm')
R_list = np.array([np.max(obj_i) / 2 for obj_i in z_obj_list])
logR_list = np.log(R_list)
R_mu_fit, R_sigma_fit = norm.fit(logR_list,
loc=R_mu_real,
scale=R_sigma_real)
R_median_fit = np.exp(R_mu_fit)
x = np.linspace(R_mu_fit - 3 * R_sigma_fit, R_mu_fit + 3 * R_sigma_fit,
1000)
pdf_fit = norm.pdf(x, R_mu_fit, R_sigma_fit)
pdf_real = norm.pdf(x, R_mu_real, R_sigma_real)
plt.figure()
plt.hist(logR_list,
bins=12,
density=True,
edgecolor='black',
linewidth=2,
color='grey',
alpha=0.5)
plt.plot(
x,
pdf_fit,
color='r',
linewidth=3.5,
label=
'empiric distribution (R_median = {0:.3f}, mu = {1:.3f}, sigma = {2:.3f})'
.format(R_median_fit, R_mu_fit, R_sigma_fit))
plt.plot(
x,
pdf_real,
color='green',
linewidth=3.5,
label=
'real distribution (R_median = {0:.3f}, mu = {1:.3f}, sigma = {2:.3f})'
.format(R_median_real, R_mu_real, R_sigma_real))
plt.xlabel(r'$\ln(R_{part}/nm)$')
plt.ylabel('frequency')
plt.title('gaussian fit, ' + r'$\sigma_{noise} = $' + str(sigma_noise) +
' nm, ' + r'$R_{tip} = $' + str(R_tip) + ' nm, ' +
r'$\frac{\mu_{fit} - \mu_{real}}{\mu_{real}} = $' +
'{:.3f}'.format((R_mu_fit - R_mu_real) / R_mu_real) +
r', $\frac{\sigma_{fit} - \sigma_{real}}{\sigma_{real}} = $' +
'{:.3f}'.format((R_sigma_fit - R_sigma_real) / R_sigma_real))
plt.legend(loc=1)
plt.tight_layout()
R_std_fit = np.sqrt((np.exp(R_sigma_fit**2) - 1) * np.exp(2 * R_mu_fit) *
np.exp(R_sigma_fit**2))
x_log = np.linspace(R_median_fit - 3 * R_std_fit,
R_median_fit + 3 * R_std_fit, 1000)
pdf_log_fit = lognorm.pdf(x_log, scale=R_median_fit, s=R_sigma_fit)
pdf_log_real = lognorm.pdf(x_log, scale=R_median_real, s=R_sigma_real)
plt.figure()
plt.hist(R_list,
bins=12,
density=True,
edgecolor='black',
linewidth=2,
color='grey',
alpha=0.5)
plt.plot(
x_log,
pdf_log_fit,
color='r',
linewidth=3.5,
label='empiric distribution (R_median = {0:.3f}, sigma = {1:.3f})'.
format(R_median_fit, R_sigma_fit))
plt.plot(
x_log,
pdf_log_real,
color='green',
linewidth=3.5,
label='real distribution (median = {0:.3f}, sigma = {1:.3f})'.format(
R_median_real, R_sigma_real))
plt.xlabel(r'$R_{part} [nm]$')
plt.ylabel('frequency')
plt.title('from fit calculated lognorm, ' + r'$\sigma_{noise} = $' +
str(sigma_noise) + ' nm, ' + r'$R_{tip} = $' + str(R_tip) +
' nm')
plt.legend(loc=1)
plt.tight_layout()
return R_median_fit, R_mu_fit, R_sigma_fit
