Drawing a directed graph using a link matrix with networkx - python

I am working on pagerank for a school project, and i have a matrix where the row "i" represent the links from the site j (line) to the site i. (If it is still unclear i'll explain more).
The current part is:
Z=[[0,1,1,1,1,0,1,0,0,0,0,0,0,0],[1,0,0,0,1,0,0,0,0,0,0,0,0,0], [1,1,0,0,0,0,0,0,0,0,0,0,0,0],[1,0,1,0,0,0,0,0,0,0,0,0,0,0],[1,0,0,1,0,0,0,0,0,0,0,0,0,0],[1,0,0,0,0,0,0,1,0,1,0,0,0,0],[0,0,0,0,0,1,0,0,0,0,0,0,0,0],[0,0,0,0,0,1,1,0,1,0,0,0,0,0],[0,0,0,0,0,1,0,0,0,0,0,0,0,0],[0,0,0,0,0,0,0,0,1,0,1,1,1,1],[0,0,0,0,0,0,0,0,0,1,0,0,0,1],[0,0,0,0,0,0,0,0,0,1,1,0,0,0],[0,0,0,0,0,0,0,0,0,1,0,1,0,0],[0,0,0,0,0,0,0,0,0,1,0,0,1,0]]
A=np.matrix(Z)
G=nx.from_numpy_matrix(A,create_using=nx.MultiDiGraph())
pos=nx.circular_layout(G)
labels={}
for i in range (N):
labels[i]=i+1
nx.draw_circular(G)
nx.draw_networkx_labels(G,pos,labels,font_size=15)
The problem i have is that the labels are not where they are supposed to be, it seems that networkx is just placing them clockwise...
Also, how could i easily direct the graph, so that a link from j to i won't be from i to j?
Thanks!


import numpy as np
import matplotlib.pyplot as plt
import networkx as nx
Z = [[0, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0]]
G = nx.from_numpy_matrix(np.array(Z), create_using=nx.MultiDiGraph())
pos = nx.circular_layout(G)
nx.draw_circular(G)
labels = {i : i + 1 for i in G.nodes()}
nx.draw_networkx_labels(G, pos, labels, font_size=15)
plt.show()
yields
This result appears correct to me. Notice, for example, that the node labeled 1 has directed edges pointing to 2, 3, 4, 5 and 7. This corresponds to the ones on the first row in the array, Z[0]:
[0, 1, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0]
since the first row corresponds to node 1, and the ones in this row occur in the columns corresponding to nodes 2, 3, 4, 5 and 7.

Related

python cv2.findContours does a bad job finding contours in a simple grayscale image

I have the following back and white image
import numpy as np
thresh = np.array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1]]).astype('uint8')
I am trying to find contour in the thresh image, like this
import cv2
contours, hierarchy = cv2.findContours(
thresh,
cv2.RETR_CCOMP,
cv2.CHAIN_APPROX_SIMPLE
)
Just looking at the threshold image, it's intuitive that there is 1 big contour around the 1, that is shaped like an arrow.
However, visually inspection of the returned contour from cv2 found using
canvas = np.zeros_like(thresh)
for ct in contours:
cv2.drawContours(canvas, ct,-1, 1, 1)
yeilds the following;
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1]],
dtype=uint8)
In summary, how would be best get the contour from thresh image?

The result you obtained is correct.
Contours allow you to find the points along the boundary of any shape. The third parameter in cv2.findContours lets you decide how you want to store the boundary points. You have 2 ways of doing that (1) either store all the points OR (2) find a good approximation.
In your case, you are using the flag cv2.CHAIN_APPROX_SIMPLE. This method does not store all the boundary points of the shape. For every line along the boundary, it stores just 2 points (the ends of each line). This is the best way to approximate the shape of any contour and it also saves memory.
If you want to store all the boundary points you need to use cv2.CHAIN_APPROX_NONE.
Here's the documentation for more
Illustration:
Consider the following array as input:
thresh = np.array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
dtype=uint8)
Using the flag cv2.CHAIN_APPROX_SIMPLE you would get:
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
dtype=uint8)
And using the flag cv2.CHAIN_APPROX_NONE you would get:
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
dtype=uint8)

Creating arrays with a loop (Python)

I am trying to create several arrays from a big array that I have. What I mean is:
data = [[0, 1, 0, 0, 0, 0, 0, 1, 0, 0], [0, 0, 1, 0, 0, 1, 0, 0, 0, 0],
[0, 1, 1, 0, 0, 0, 0, 0, 0, 1], [1, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 0, 1, 1, 0, 0, 0, 0, 0,1], [0, 0, 0, 0, 1, 1, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 1, 0, 0], [0, 1, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 1, 0]]
I want to create 10 different arrays - using the 10 data's columns - with different names.
data1 = [0, 0, 0, 1, 0, 0, 1, 0, 0],
data2 = [1, 0, 1, 0, 0, 0, 0, 1, 0], and so on
I found a close solution here - Also I take the example data from there - However, when I tried the solution suggested:
for d in xrange(0,9):
exec 'x%s = data[:,%s]' %(d,d-1)
A error message appears:
exec(code_obj, self.user_global_ns, self.user_ns)
File "", line 2, in
exec ('x%s = data[:,%s]') %(d,d-1)
File "", line 1
x%s = data[:,%s]
^
SyntaxError: invalid syntax
Please, any comments will be highly appreciated. Regards

Use numpy array index:
data = [[0, 1, 0, 0, 0, 0, 0, 1, 0, 0], [0, 0, 1, 0, 0, 1, 0, 0, 0, 0],
[0, 1, 1, 0, 0, 0, 0, 0, 0, 1], [1, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 0, 1, 1, 0, 0, 0, 0, 0,1], [0, 0, 0, 0, 1, 1, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 1, 0, 0], [0, 1, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 1, 0]]
d = np.array(data)
d[:, 0]
#array([0, 0, 0, 1, 0, 0, 1, 0, 0])
d[:, 1]
#array([1, 0, 1, 0, 0, 0, 0, 1, 0])
etc...
d[:, 9]
#array([0, 0, 1, 1, 1, 0, 0, 0, 0])
If you must, then dictionaries are the way to go:
val = {i:d[:,i] for i in range(d.shape[1])}
To access the arrays:
val[0]
#array([0, 0, 0, 1, 0, 0, 1, 0, 0])
...
val[9]
#array([0, 0, 1, 1, 1, 0, 0, 0, 0])

Use the following code (it is also more readable -- for python 3.x) if you really want to create dynamic variables:
for d in range(0,9):
# exec 'x%s = data[:,%s]' %(d,d-1)
exec(f"data{d} = {data[d]}" )

Either use numpy array as shown by scott boston above or use dictionary like this:
a = {}
for i in range(0,9):
a[i] = data[i][:]
Output:
{0: [0, 1, 0, 0, 0, 0, 0, 1, 0, 0],
1: [0, 0, 1, 0, 0, 1, 0, 0, 0, 0],
2: [0, 1, 1, 0, 0, 0, 0, 0, 0, 1],...

I don't see the proper indentation in your for loop.
I suggest you don't use %s for the second argument (string) but rather %d (number) since you need a number to do the indexing of your array.

How to solve 4 x 4 array with 16 unknown with Python (Update)

I need to solve a 2 by 2 array with 4 unknown
A B
C D
I know all horizontal sum A+B=11, C+D=7
I know all vertical sum A+C=10, B+D=8
I know all diagonal sum A+D=15, B+C=3
I then use Python to solve for A,B,C,D
import numpy as np
A = [[1, 1, 1, 1],
[1, 0, 0, 1],
[1, 0, 1, 0],
[0, 0, 1, 1]]
a = [18, 15, 10, 7]
answera = np.linalg.solve(A, a)
print(answera)
And the answer is [9. 2. 1. 6.] which is correct
Now I need to solve 4 by 4 array with 16 unknown
A B C D
E F G H
I J K L
M N O P
I know horizontal sum A+B+C+D=10, E+F+G+H=26, I+J+K+L=42, M+N+O+P=58
I know vertical sum A+E+I+M=28, B+F+J+N=32, C+G+K+O=36, D+H+L+P=40
I know diagonal sum M=13, I+N=23, E+J+O=30, A+F+K+P=34, B+G+L=21, C+H=11, D=4
The other diagonal sum A=1, B+E=7, C+F+I=18, D+G+J+M=34, H+K+N=33, L+O=27, P=16
Which mean I know the value of the 4 corners.
I tried the following code but did not work
C = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1],
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0],
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
c = [10, 26, 42, 58, 7, 21, 39, 33, 27, 11, 23, 35, 30, 23, 32, 136]
answerc = np.linalg.solve(C, c)
print(answerc)
The correct answer should be [1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.] but I got error message
Traceback (most recent call last):
answerc = np.linalg.solve(C, c)
r = gufunc(a, b, signature=signature, extobj=extobj)
raise LinAlgError("Singular matrix")
numpy.linalg.LinAlgError: Singular matrix
Am I in the right direction? I will need to solve 5X5 with 25 unknown, 6X6 with 36 unknown and so on. Is there an easier way?
-----------------------------------------------------------------------------
Following Mr. Rory Daulton solution, I can solve the above 1 to 16 4X4 array without problem, but when I use it in another array with negative number, it doesn't give answer as expected;
The negative 4X4 array as follow
-20 -10 -5 0
-10 -20 -10 -5
-5 0 -10 -20
-10 -20 -10 -5
My python code as follow
import numpy as np
G = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], # horizontal rows
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1],
[1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0], # vertical columns
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0],
[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0], # forward diagonals
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], # back diagonals
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
]
g = [-35, -45, -35, -45, # horizontal rows
-45, -50, -35, -30, # vertical columns
-10, -25, -20, -55, -40, -10, 0, # forward diagonals
-20, -20, -30, -20, -35, -30, -5, # back diagonals
]
answerg = np.linalg.lstsq(G, g, rcond=None)
print(answerg[0])
The output is not exactly the original array
[-2.00000000e+01 -1.31250000e+01 -1.87500000e+00 8.88178420e-15
-6.87500000e+00 -2.00000000e+01 -1.00000000e+01 -8.12500000e+00
-8.12500000e+00 2.13162821e-14 -1.00000000e+01 -1.68750000e+01
-1.00000000e+01 -1.68750000e+01 -1.31250000e+01 -5.00000000e+00]
What should I try? Thank you in advance.

SHORT ANSWER: There are infinitely many solutions to your problem. So this takes a more complex analysis of the equations.
LONG ANSWER: You have multiple problems with your code.
First, you make it easy to make mistakes, since the lines of your matrix do not correspond to the data that you present. Worse, you have no comments to explain things. This mis-match will probably cause errors. You have 22 pieces of data in your sums, so use them. You tried to combine some of the sums and ignore others (the four corners) but you did not do it properly and you ended up with a singular matrix.
Next, you use linalg.solve. In your problem you have more data items (22) than unknowns (16), so solve is inappropriate. The numpy documentation for solve states
a must be square and of full-rank, i.e., all rows (or, equivalently,
columns) must be linearly independent; if either is not true, use
lstsq for the least-squares best “solution” of the system/equation.
The matrix resulting from your data is not square, therefore the rows are not linearly independent, so you should use lstsq rather than solve. The lstsq routine gives more information than you need for your problem, so just print the first item in the resulting list.
Combining those ideas and adding a few comments gives this code:
import numpy as np
C = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], # horizontal rows
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1],
[1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0], # vertical columns
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0],
[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0], # forward diagonals
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], # back diagonals
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
]
c = [10, 26, 42, 58, # horizontal rows
28, 32, 36, 40, # vertical columns
13, 23, 30, 34, 21, 11, 4, # forward diagonals
1, 7, 18, 34, 33, 27, 16, # back diagonals
]
answerc = np.linalg.lstsq(C, c, rcond=None)
print(answerc[0])
The printout is what you want:
[ 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.]
However, to be honest, there is no guarantee that this is an answer--just that it is a "closest" answer. Also, if it is an answer, there may be other answers. And, in fact, further analysis shows that there are other answers that satisfy all your conditions.
The sympy module can generate a row reduced echelon form of the matrix, which can be used to do more in-depth analysis of all the answers. However, the constants are then to be part of the matrix, rather than used as a separate array. Here is code for sympy to attempt to solve your problem:
import sympy
C = [[1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 10], # horizontal rows
[0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 26],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 42],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 58],
[1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 28], # vertical columns
[0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 32],
[0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 36],
[0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 40],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 13], # forward diagonals
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 23],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 30],
[1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 34],
[0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 21],
[0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 11],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4],
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1], # back diagonals
[0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7],
[0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 18],
[0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 34],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 33],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 27],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 16],
]
print(sympy.Matrix(C).rref())
The printout is
(Matrix([
[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
[0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, -1, 0, -13],
[0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 18],
[0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4],
[0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 20],
[0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6],
[0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7],
[0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, -1, 0, -7],
[0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, -1, 0, -6],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 10],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 11],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 27],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 13],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 29],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 16],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15))
If you know how to analyze this you will see that there are infinitely many answers to your problem. If you set the number in the 15th cell to 15+x then the above matrix shows that the answer to all your restrictions is
1 2+x 3-x 4
5-x 6 7 8+x
9+x 10 11 12-x
13 14-x 15+x 16
The solve function of numpy only works if there is just one solution, so even if you had adjusted your matrix differently it would not have worked for you.
ANSWER TO YOUR UPDATE:
It seems that you missed the point of my answer. Your 4x4 problem has infinitely many answers, so there is no procedure that can choose the particular answer that you have in mind. The np.linalg.lstsq routine can find one of the answers to your problem but probably will not find your desired answer. You should consider it to be a coincidence that using that routine in your first problem gave your desired answer--that will probably not work in other problems.
It is a little hard to interpret the given answer to your new problem, since the scientific notation is hard to read. But all those matrix values are exact, and here they are as rational numbers in a particular format that should be obvious:
-20 -10-(3+1/8) - 5+(3+1/8) 0
-10+(3+1/8) -20 -10 - 5-(3+1/8)
- 5-(3+1/8) 0 -10 -20+(3+1/8)
-10 -20+(3+1/8) -10-(3+1/8) - 5
You see that the numpy's answer is the one that you expected, with the value 3+1/8 added to or subtracted from half the array values. This makes x=3+1/8 in the general answer that I gave you for your first problem.
This is as good as you can expect. Numpy gave you a correct answer--it has no idea how to choose the answer that you had in your head from the infinitely many correct answers to your problem. The only way to get just one answer is to change your problem--perhaps state the value in the first row and second column, or the sum of the first and third values in any one of the rows, or something similar.

how to plot Multidimensional list (list of lists ) in matplot?

how to plot Multidimensional list like one below using matplot
d=[[1 ,1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0, 0, 0, 1, 2, 2, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1 ,1 ,0, 0, 1, 2, 2, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0],
[1 ,0 ,1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0 ,0 ,0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 1, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 1, 0, 0, 0, 0]]
i used KMeans algorithm from scikitlearn to form clusters but i need a way to visualize these clusters.
here my code for k-means
cl= KMeans(n_clusters=4)
cl.fit(d)
cen = cl.cluster_centers_
lab=cl.labels_
i know how to use matplot to plot simple graph but i never used to Multidimensional. i want to plot d and then cluster_centers (centroids)
plot should be something like this
is there any hit may help me accomplish this task?

As #Mahdi said, you have to use some dimensionality reduction to be able to plot the points from high-dimensional space on two-dimensional screen. (Of course, you will loose some information, but that's unavoidable.) Here is an example to how to do it with PCA (however, there are different techniques, check the references above).
from sklearn.decomposition import PCA
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
%matplotlib inline
# d = ... (the same as in the question)
n_clusters = 4
cl= KMeans(n_clusters=n_clusters)
cl.fit(d)
cen = cl.cluster_centers_
lab=cl.labels_
pca = PCA(2)
# reduce dimension to 2
coords = pca.fit_transform(d)
plt.scatter(coords[:, 0], coords[:, 1], 20, c=lab, cmap='spring')
centroids_transformed = pca.transform(cen)
# transform centroids coordinate to new space
plt.scatter(centroids_transformed[:, 0], centroids_transformed[:, 1], 60,
c=range(n_clusters), cmap='spring', marker='v')
# plot centroids. they are presented in the same order as labels,
# so colors are just range

Python Equivalent for bwmorph

I am still coding a fingerprint image preprocessor on Python. I see in MATLAB there is a special function to remove H breaks and spurs:
bwmorph(a , 'hbreak')
bwmorph(a , 'spur')
I have searched scikit, OpenCV and others but couldn't find an equivalent for these two use of bwmorph. Can anybody point me to right direction or do i have to implement my own?

Edit October 2017
the skimage module now has at least 2 options:
skeletonize and thin
Example with comparison
from skimage.morphology import thin, skeletonize
import numpy as np
import matplotlib.pyplot as plt
square = np.zeros((7, 7), dtype=np.uint8)
square[1:-1, 2:-2] = 1
square[0, 1] = 1
thinned = thin(square)
skel = skeletonize(square)
f, ax = plt.subplots(2, 2)
ax[0,0].imshow(square)
ax[0,0].set_title('original')
ax[0,0].get_xaxis().set_visible(False)
ax[0,1].axis('off')
ax[1,0].imshow(thinned)
ax[1,0].set_title('morphology.thin')
ax[1,1].imshow(skel)
ax[1,1].set_title('morphology.skeletonize')
plt.show()
Original post
I have found this solution by joefutrelle on github.
It seems (visually) to give similar results as the Matlab version.
Hope that helps!
Edit:
As it was pointed out in the comments, I'll extend my initial post as the mentioned link might change:
Looking for a substitute in Python for bwmorph from Matlab I stumbled upon the following code from joefutrelle on Github (at the end of this post as it's very long).
I have figured out two ways to implement this into my script (I'm a beginner and I'm sure there are better ways!):
1) copy the whole code into your script and then call the function (but this makes the script harder to read)
2) copy the code it in a new python file 'foo' and save it. Now copy it in the Python\Lib (eg. C:\Program Files\Python35\Lib) folder. In your original script you can call the function by writing:
from foo import bwmorph_thin
Then you'll feed the function with your binary image:
skeleton = bwmorph_thin(foo_image, n_iter = math.inf)
import numpy as np
from scipy import ndimage as ndi
# lookup tables for bwmorph_thin
G123_LUT = np.array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1,
0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 0,
1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,
0, 1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 1,
0, 0, 0], dtype=np.bool)
G123P_LUT = np.array([0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0,
1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0,
0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0,
1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 1,
0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0], dtype=np.bool)
def bwmorph_thin(image, n_iter=None):
"""
Perform morphological thinning of a binary image
Parameters
----------
image : binary (M, N) ndarray
The image to be thinned.
n_iter : int, number of iterations, optional
Regardless of the value of this parameter, the thinned image
is returned immediately if an iteration produces no change.
If this parameter is specified it thus sets an upper bound on
the number of iterations performed.
Returns
-------
out : ndarray of bools
Thinned image.
See also
--------
skeletonize
Notes
-----
This algorithm [1]_ works by making multiple passes over the image,
removing pixels matching a set of criteria designed to thin
connected regions while preserving eight-connected components and
2 x 2 squares [2]_. In each of the two sub-iterations the algorithm
correlates the intermediate skeleton image with a neighborhood mask,
then looks up each neighborhood in a lookup table indicating whether
the central pixel should be deleted in that sub-iteration.
References
----------
.. [1] Z. Guo and R. W. Hall, "Parallel thinning with
two-subiteration algorithms," Comm. ACM, vol. 32, no. 3,
pp. 359-373, 1989.
.. [2] Lam, L., Seong-Whan Lee, and Ching Y. Suen, "Thinning
Methodologies-A Comprehensive Survey," IEEE Transactions on
Pattern Analysis and Machine Intelligence, Vol 14, No. 9,
September 1992, p. 879
Examples
--------
>>> square = np.zeros((7, 7), dtype=np.uint8)
>>> square[1:-1, 2:-2] = 1
>>> square[0,1] = 1
>>> square
array([[0, 1, 0, 0, 0, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0]], dtype=uint8)
>>> skel = bwmorph_thin(square)
>>> skel.astype(np.uint8)
array([[0, 1, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0]], dtype=uint8)
"""
# check parameters
if n_iter is None:
n = -1
elif n_iter <= 0:
raise ValueError('n_iter must be > 0')
else:
n = n_iter
# check that we have a 2d binary image, and convert it
# to uint8
skel = np.array(image).astype(np.uint8)
if skel.ndim != 2:
raise ValueError('2D array required')
if not np.all(np.in1d(image.flat,(0,1))):
raise ValueError('Image contains values other than 0 and 1')
# neighborhood mask
mask = np.array([[ 8, 4, 2],
[16, 0, 1],
[32, 64,128]],dtype=np.uint8)
# iterate either 1) indefinitely or 2) up to iteration limit
while n != 0:
before = np.sum(skel) # count points before thinning
# for each subiteration
for lut in [G123_LUT, G123P_LUT]:
# correlate image with neighborhood mask
N = ndi.correlate(skel, mask, mode='constant')
# take deletion decision from this subiteration's LUT
D = np.take(lut, N)
# perform deletion
skel[D] = 0
after = np.sum(skel) # coint points after thinning
if before == after:
# iteration had no effect: finish
break
# count down to iteration limit (or endlessly negative)
n -= 1
return skel.astype(np.bool)
"""
# here's how to make the LUTs
def nabe(n):
return np.array([n>>i&1 for i in range(0,9)]).astype(np.bool)
def hood(n):
return np.take(nabe(n), np.array([[3, 2, 1],
[4, 8, 0],
[5, 6, 7]]))
def G1(n):
s = 0
bits = nabe(n)
for i in (0,2,4,6):
if not(bits[i]) and (bits[i+1] or bits[(i+2) % 8]):
s += 1
return s==1
g1_lut = np.array([G1(n) for n in range(256)])
def G2(n):
n1, n2 = 0, 0
bits = nabe(n)
for k in (1,3,5,7):
if bits[k] or bits[k-1]:
n1 += 1
if bits[k] or bits[(k+1) % 8]:
n2 += 1
return min(n1,n2) in [2,3]
g2_lut = np.array([G2(n) for n in range(256)])
g12_lut = g1_lut & g2_lut
def G3(n):
bits = nabe(n)
return not((bits[1] or bits[2] or not(bits[7])) and bits[0])
def G3p(n):
bits = nabe(n)
return not((bits[5] or bits[6] or not(bits[3])) and bits[4])
g3_lut = np.array([G3(n) for n in range(256)])
g3p_lut = np.array([G3p(n) for n in range(256)])
g123_lut = g12_lut & g3_lut
g123p_lut = g12_lut & g3p_lut
"""`

You will have to implement those on your own since they aren't present in OpenCV or skimage as far as I know.
However, it should be straightforward to check MATLAB's code on how it works and write your own version in Python/NumPy.
Here is a guide describing in detail NumPy functions exclusively for MATLAB users, with hints on equivalent functions in MATLAB and NumPy:
Link

Categories