palsma-image

Convenient Image Data Analytics for Plasma and Astrophysics Research

This library provides algorithms for data analysis and image processing, including automation of feature extraction and characterization of image regions.

Example Outputs

Results of processing a synthesized image that represents peaks and valleys:

Number of detected peaks: 11
Number of detected regions around the peaks: 11

Regions Characterized:
Region 1: Avg. Thickness: 3.0572496387314114 ; Length: 20.195087901804587
Region 2: Avg. Thickness: 5.045338744098684 ; Length: 14.345840282578262
...

Detected peak points	Detected regions around the peak points

Results of processing an image obtained from a plasma simlation:

Number of detected peaks (maximas): 54
Number of detected regions (current sheets) around the peaks: 54

Regions Characterized:
Current sheet 1: Avg. Thickness: 0.4704377263922215 ; Length: 8.961775377533403
Current sheet 2: Avg. Thickness: 0.799345244773368 ; Length: 10.27921886368474
...

Detected peak points	Detected regions (current sheets) around the peak points

Data filtering and smoothing:

• We implemented Svizky-Gulay which is one of the digital filtering methods (convolution process). In this method, the local data point is fitted by the sub-set of adjacent data point sequentially with low degree polynomial by the linear least method.
  

  Here we implemented the data filtering for diffrent values of n_f (degree of polynomal) and n_w (grid point) and same time frame of simulation .

Cropping out and characterizing some of the detected regions (the line represents length):

Region 6	Region 10	Region 46

Example of statistical analysis:

In the below figures, statistical analysis of extracted features (current sheets) and distribution of size geometrical properties(thickness, length) from the simulation are shown.

Note

This library can be also beneficial to the astrophysics society. For example you can perform statistical analysis of thickness, length, and aspect ratio (length/half-thickness) of each current-sheet in a plasma. Examples of visualizations and results can be seen in a publication in the Journal of Physics of Plasmas: doi.org/10.1063/5.0040692

Contributing

Pull requests are welcome!

Installation

pip install scimage

Using the library

See examples (simple-example.py and example.py) in the github repository, on how to use the library functions. Before running the examples, make sure to download the data files from the data folder of the repository.

Simple Example:

import sys
import numpy as np
import matplotlib.pyplot as plt

import scimage.identification as ident # Identification of peaks and regions (image segmentation)
import scimage.characterization as char # Characterizing each detected region (e.g. thickness and length)
import scimage.plot_functions as scplt # Plotting functions
from scimage.file_functions import (load_simulation_image)

sys.setrecursionlimit(10000) # to avoid possible RecursionError


# Prepare a 2D plane image
values, nx, ny, lx, ly, coordinates_x, coordinates_y = load_simulation_image('data/data-512-512.npz')

noise_threshold = 0.1
ratio_of_boundary_to_max = 0.5
points_upto_local_boundary = 10

values_abs = np.abs(values)

# Detect peak points (local maximas):
good_indexes = ident.remove_noise(values_abs, noise_threshold)
indexes_of_peaks, peak_values, array_with_peaks_only = \
    ident.find_local_maxima_at_selected_indexes(values_abs, good_indexes, points_upto_local_boundary)

# Detect regions surrounding each maxima point (image segmentation)
indexes_of_points_of_all_regions, indexes_of_valid_peaks = \
    ident.detect_regions_around_peaks(values_abs, indexes_of_peaks, ratio_of_boundary_to_max)

print ("Number of detected peaks:" , len(indexes_of_peaks))
print ("Number of detected regions around the peaks:" , len(indexes_of_points_of_all_regions))

# Plot the whole image plane, together with the detected peaks and regions:
plt.rcParams["figure.autolayout"] = True # Enable tight layout with minimum margins
plt.rcParams["figure.figsize"] = (10, 8) # Set the desired figure size

plt.ioff()
scplt.plot_locations_of_local_maximas(coordinates_x, coordinates_y, values, noise_threshold, indexes_of_peaks)
scplt.plot_locations_of_region_points(coordinates_x, coordinates_y, values, noise_threshold, indexes_of_points_of_all_regions)
plt.show() # Show the plots. This also pauses the script here so that we can see the plots


# Characterize one of the detected regions -------------------------------
selected_region = 0 # choose one region as an example
indexes_of_points_of_one_region = indexes_of_points_of_all_regions[selected_region]

# First, cut out the selected region as a separate frame from the whole image
coordinates_x_in_frame, coordinates_y_in_frame, values_of_frame = \
    char.build_region_frame(indexes_of_points_of_one_region, coordinates_x, coordinates_y, values)

# Now, estimate thickness of the region
min_val = np.max(values_of_frame) * 0.42
half_thickness_plus_side, half_thickness_minus_side = \
    char.characterize_region(values_of_frame, coordinates_x_in_frame, coordinates_y_in_frame, min_val)

# Also, find length with the pair-wise comparison method
length, p1, p2 = char.find_length_by_pariwise_distance(indexes_of_points_of_one_region, coordinates_x, coordinates_y)

print()
print("Region", selected_region,"with frame size in pixels", values_of_frame.shape, "characterized:")
print("\tLength:", length)
print("\tThickness (half plus, half minus):", half_thickness_plus_side, half_thickness_minus_side)


# Plot one region and save its image
plt.rcParams["figure.figsize"] = (4, 4) # Set the desired figure size
plt.ioff()
scplt.plot_region(coordinates_x_in_frame, coordinates_y_in_frame, values_of_frame, p1, p2, region_index = selected_region)
plt.show() # Show the plots. This also pauses the script here so that we can see the plots