def get_mask(samples, backgrounds=None, xoff=None, yoff=None, cfg=None):
""" Returns a thresholded mask where an imaged object is suspected to be.
It could be useful in high contrast images where object borders are
sufficiently well determined.
Args:
-----
samples: the acquired samples as a dictionary with angles as keys.
backgrounds: the acquired background as a dictionary with angles as
keys. They must be acquired right after or before taking
the samples.
xoff: offset in the 'x' direction of this sample's center.
yoff: offset in the 'y' direction of this sample's center.
cfg: configuration (named tuple)
Returns:
--------
(ndarray) mask containing the object to be reconstructed.
"""
corr_ims = list()
iterator = fpmm.set_iterator(cfg)
for index, theta, shift in iterator:
image = samples[(theta, shift)]
image = fpmm.crop_image(image, cfg.patch_size, xoff, yoff)
# image, image_size = image_rescaling(image, cfg)
background = backgrounds[(theta, shift)]
background = fpmm.crop_image(background, cfg.patch_size, xoff, yoff)
# background, image_size = image_rescaling(background, cfg)
corr_ims.append(image_correction(image, background, mode='background'))
mask = np.mean(corr_ims, axis=0)
#
thres = 140 # hardcoded
mask[mask < thres] = 1
mask[mask > thres] = 0
# print(Et[np.abs(Et) > .1])
return mask
def initialize(hrsize=None,
backgrounds=None,
xoff=None,
yoff=None,
cfg=None,
mode='zero'):
""" Initializes the algorithm using one of various modalities.
Args:
-----
samples: the acquired samples as a dictionary with angles as keys.
backgrounds: the acquired background as a dictionary with angles as
keys. They must be acquired right after or before taking
the samples.
xoff: offset in the 'x' direction of this sample's center.
yoff: offset in the 'y' direction of this sample's center.
cfg: configuration (named tuple)
mode: the modality according to which the algorithm is going to be
initilized. Can be one of the following options:
* zero: constant phase and zero amplitude.
* transmission: the transmitted image at (0, 0) angles.
* mean: takes all the samples and substracts the (measured)
backround. Then takes the mean of all of them.
Returns:
--------
(ndarray) a complex image with the same size of the samples.
"""
# image = fpmm.crop_image(samples[(0, 0)], cfg.patch_size, xoff, yoff)
# image, image_size = image_rescaling(image, cfg)
if mode == 'zero':
Ih_sq = 0.5 * np.ones(hrsize) # Homogeneous amplitude
Ph = np.zeros(hrsize) # and null phase
Et = Ih_sq * np.exp(1j * Ph)
elif mode == 'transmission':
background = fpmm.crop_image(backgrounds[(0, 0)], cfg.patch_size, xoff,
yoff)
background, image_size = image_rescaling(background, cfg)
image = image_correction(image, background, mode='background')
Ih_sq = np.sqrt(image)
Ph = np.zeros_like(Ih_sq) # and null phase
Et = Ih_sq * np.exp(1j * Ph)
elif mode == 'mean':
corr_ims = list()
iterator = fpmm.set_iterator(cfg)
for index, theta, shift in iterator:
# Final step: squared inverse fft for visualization
image = fpmm.crop_image(samples[(theta, shift)], cfg.patch_size,
xoff, yoff)
background = fpmm.crop_image(backgrounds[(theta, shift)],
cfg.patch_size, xoff, yoff)
image, image_size = image_rescaling(image, cfg)
background, image_size = image_rescaling(background, cfg)
corr_ims.append(
image_correction(image, background, mode='background'))
Ih = np.mean(corr_ims, axis=0)
# Ph = 0.5+np.pi*np.abs(Et)/np.max(Et)
Et = np.sqrt(Ih) * np.exp(1j * 0)
return Et
# from pyfpm.data import json_savemeta, json_loadmeta
import pyfpm.data as dt
from pyfpm.coordinates import PlatformCoordinates
from pyfpm import implot
import pyfpm.local as local
# Simulation parameters
CONFIG_FILE = './etc/config.yaml'
cfg = dt.load_config(CONFIG_FILE)
simclient = local.SimClient(cfg=cfg)
pc = PlatformCoordinates(theta=0, phi=0, height=cfg.sample_height, cfg=cfg)
in_file = os.path.join(cfg.output_sample, '2017-04-12_182910.npy')
ss_dict = np.load(cfg.output_sample+'2017-04-12_182910ss_dict.npy')[()]
power_dict = np.load(cfg.output_sample+'2017-04-12_182910power_dict.npy')[()]
iterator = fpm.set_iterator(cfg)
image_dict = np.load(in_file)[()]
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=(25, 15))
plt.grid(False)
fig.show()
cum_diff = 0
for index, theta, phi in iterator:
# Coordinates math
# time.sleep(0.5)
pc.heigth = 88
pc.platform_tilt = [0, 2]
pc.source_center = [0, 2] #[xcoord, ycoord] of the calibrated center
pc.set_coordinates(theta=theta, phi=phi, units='degrees')
t_corr, p_corr = pc.source_coordinates(mode='angular')
from pyfpm import web
from pyfpm.reconstruct import fpm_reconstruct
from pyfpm.fpmmath import set_iterator
from pyfpm.data import save_yaml_metadata
import pyfpm.data as dt
from pyfpm.coordinates import PlatformCoordinates
# Simulation parameters
cfg = dt.load_config()
samples, comp_cfg = dt.open_sampled('2017-05-26_152307.npy')
background, comp_cfg = dt.open_sampled('2017-05-26_145449_blank.npy')
# Connect to a web client running serve_microscope.py
pc = PlatformCoordinates(theta=0, phi=0, height=cfg.sample_height, cfg=cfg)
pc.generate_model(cfg.plat_model)
iterator = set_iterator(cfg)
# reconstruction
dx = cfg.patch_size[0]
x_range = range(0, cfg.video_size[1], dx)[:-1]
y_range = range(0, cfg.video_size[0], dx)[:-1]
# for i, point in enumerate(it.product(x_range, y_range)):
# init_point = [point[0], point[1]]
# rec, phase = fpm_reconstruct(samples, background, iterator, init_point,
# cfg=cfg, debug=False)
# misc.imsave('./misc/ph'+str(i)+'.png', phase)
# misc.imsave('./misc/im'+str(i)+'.png', rec)
rec, phase = fpm_reconstruct(samples, background, iterator, [90, 80],
cfg=comp_cfg, debug=True)
plt.show()