import numpy as np
from ..img.image import Image
from ..fun.misc import reduce_list as rl
[docs]
class Gauss2D(Image):
"""Plot Gaussian spots according to X,Y-coordinate.
.. caution::
Trajectory data should be converted to localization data.
Args:
reqs[0] (Table): Localization data including X,Y-coordinate columns.
Required param; ``length_unit``.
param["pitch"] (float): Length per pixel of reconstructed image.
param["sd"] (float): Standard deviation of Gaussian spots.
param["img_size] (list of int): Image size as [width, height] pixels.
param["window_factor"] (float): S.D. of Gaussian spots is multiplied by
this factor for the half pixel width of rendering clip of Gaussian
spots. If you set a too small value, Gaussian spots will be
truncated square-shaped spots.
param["group_depth"] (int): Plots grouping depth. The grouped
coordinates by this depth will be plotted into the same frame.
param["split_depth"] (int): File split depth number.
Returns:
Image: Image class of reconstructed ``float32`` tiff file
"""
[docs]
def set_info(self, param={}):
"""Copy info from reqs[0] and modify columns and add params.
"""
self.info.copy_req(0)
length_unit = self.info.get_param_value("length_unit")
self.info.set_group_depth(param["group_depth"])
self.info.delete_column(keeps=self.info.get_param_value("index_cols"))
self.info.add_column(
0, "intensity", "float32", "a.u.", "Pixel intensity")
self.info.add_param(
"calc_cols", ["x_" + length_unit, "y_" + length_unit],
"list of str", "X,Y-coordinate columns")
self.info.add_param(
"pitch", param["pitch"], length_unit + "/pix",
"Length per pixel of reconstructed image")
self.info.add_param(
"img_size", param["img_size"], "pix",
"Image size as [width, height] pixels")
self.info.add_param(
"sd", param["sd"], length_unit, "Standard deviation of Gauss plot")
self.info.add_param(
"window_factor", param["window_factor"],
"float", "Multiplying factor of plot S.D. for half window width")
self.info.set_split_depth(param["split_depth"])
[docs]
@ staticmethod
def process(reqs, param):
"""Plot Gaussian spots according to X,Y-coordinate.
Args:
reqs[0] (pandas.DataFrame): Localization data including
X,Y-coordinate columns.
param["calc_cols"] (list of str): X,Y-coordinate columns.
param["pitch"] (float): Length per pixel of reconstructed image.
param["sd"] (float): Standard deviation of Gaussian spots.
param["img_size] (list of int): Image size as [width, height]
pixels.
param["window_factor"] (float): S.D. of Gaussian spots is
multiplied by this factor for the half pixel width of rendering
clip of Gaussian spots. If you set a too small value, Gaussian
spots will be truncated square-shaped spots.
param["index_cols"] (list of str): Column names of index.
Returns:
numpy.ndarray: Reconstructed image stack
"""
df = reqs[0].copy()
width_unit = param["img_size"][0]
width = np.floor(width_unit).astype(np.int32)
height_unit = param["img_size"][1]
height = np.floor(height_unit).astype(np.int32)
grouped = df.groupby(rl(param["index_cols"]))
img = []
for _, df_group in grouped:
xcs = df_group[param["calc_cols"][0]].values / param["pitch"]
ycs = df_group[param["calc_cols"][1]].values / param["pitch"]
frm = np.zeros([height, width])
sd = param["sd"] / param["pitch"]
window = np.ceil(sd * param["window_factor"])
x_pix = np.arange(0.5, 2 * window + 1, 1.0)
y_pix = np.arange(0.5, 2 * window + 1, 1.0)
x_grid, y_grid = np.meshgrid(x_pix, y_pix)
for xc, yc in zip(xcs, ycs):
yc = width - yc
dx = xc - np.floor(xc)
left = (xc - dx - window).astype(int)
right = (xc - dx + window + 1).astype(int)
dy = yc - np.floor(yc)
top = (yc - dy - window).astype(int)
bottom = (yc - dy + window + 1).astype(int)
if np.all([0 <= left, right <= width, 0 <= top, bottom
<= width]):
add_vals = gauss2d(
x_grid, y_grid, window + dx, window + dy, sd)
frm[top:bottom, left:right] = \
frm[top:bottom, left:right] + add_vals
frm = np.flipud(frm)
img.append(frm)
return np.array(img)
[docs]
def gauss2d(x, y, xc, yc, s):
"""Equation for plotting 2D Gaussian image.
Args:
x (float): X-coordinate of grid position.
y (float): Y-coordinate of grid position.
xc (float): X center of Gaussian distribution.
yc (float): Y center of Gaussian distribution.
s (float): Standard deviation of Gaussian distribution.
Returns:
float: Gaussian value of (x, y)
"""
return 1. / (2. * np.pi * s**2) * np.exp(-((x - xc)**2. + (y - yc)**2.) /
(2. * s**2.))