How to do PCA for EELS SI in python

PCA is the popular statistic tool to denoise the EELS spectrum image. Here is the simple example to apply PCA by numpy package

==================================

"""

ref:

https://stackoverflow.com/questions/13224362/principal-component-analysis-pca-in-python/49629816#49629816

@author: renfong

"""

import numpy as np

import matplotlib.pyplot as plt

import hyperspy.api as hs

from numpy import argsort

from numpy.linalg import eigh

#%% load data

data = hs.load('t1.dm3').data

#%% define pca

def pca(data, pc_count = None):

"""

Principal component analysis using eigenvalues

note: this mean-centers and auto-scales the data (in-place)

"""

C = np.dot(data.T, data) # covariance matrix

E, V = eigh(C)

key = argsort(E)[::-1][:pc_count]

E, V = E[key], V[:, key]

U = data @ V # equvalent to np.dot(data, V)

return U, E, V

#%% reconstruction

U, _, V = pca(data, 3)

recons = U @ V.T

""" visualize """

ch = 80

plt.figure()

plt.plot(data[ch,:], label='Raw')

plt.plot(recons[ch,:], label='PCA')

plt.legend()

plt.xlim([800,1800])

plt.ylim([data[ch,1850], data[ch,750]])

plt.show()

====

https://github.com/Renfong/Python-toolbox/blob/main/PCA_numpy.py

Comments

Top hat filter

The top_hat filter can be used to detect the relatively small edges/peaks superimposed on large background signals. The concept came from the EELS workshop during IMC19. Thanks to Prof. Nestor J. Zaluzec. -- // Using Top_hat digital filter to detect the relatively small edges // superimposed on large background signals. // // ref: Ultramicroscopy 18 (1985) 185-190 // Digital Filters for Application to Data Analysis in EELS // by Nestor J. ZALUZEC // Parameters: // win_s: signal window (default:3) // win_b: background window (default:3) // a_s : amplitude of signal (fixed value) // a_b : amplitude of background (fixed value) // Renfong 2018/10/11 // Main function image Top_Hat_Filter(image img, number win_s, number win_b) { // read image string fname=img.GetName() number sx,sy img.getsize(sx,sy) // filter image img2 := imageclone(img)*0 //the area between...

HyperSpy - read the calibration information in a dm3/dm4 file

Some example of dm3 file reading by using Python HyperSpy package, which can read the detail information of the dm file. -- # import packages import numpy as np import hyperspy.api as hs # load file sp=hs.load('sp.dm3') # Read the axis information # Print all the calibration detail print(sp.axes_manager) ''' <Axes manager, axes: (272|2042)> Name | size | index | offset | scale | units ================ | ======= | ====== | ======= | ======= | ====== x | 272 | 0 | -0 | 0.0025 | µm --------------- | ------ | ----- | ------ | ------- | ------ Energy loss | 2042 | | 3.2e+02 | 1 | eV...

MLLS in matlab

MLLS stands for multiple linear least squares fitting, which is the common strategy for the solving EELS edge overlapping and which is also built-in the GMS software. The target spectrum Y and the reference spectrum X Y = A * X Assuming Y is 1*256 matrix and we have three reference spectrums, ie, X is 3*256 matrix. So A is 1*3 matrix. The target is to solve A. If Y and X are n*n matrices, we can use the simple formula Y * inv(X) = A * X * inv(X), ie., A = Y * inv(X). However, Y and X are not n*n matrices, it is necessary to have some trick to solve it. We can multiply the transpose matrix to produce n*n matrix. Y * X' = A * X * X' (ps X' means the transpose matrix of X) so A = Y * X' * inv(X * X') Here is the Matlab code: ========= % create target spectrum x=0:256; c=[90,120,155]; sig=[5,10,8]; int=[5,10,8]; xn=zeros(size(x)); ref=zeros(length(c),length(x)); factor=rand(size(c))'; for i=1:length(c) xn=xn+int(i)*ex...

Renfong's Page

Search This Blog