Permanent Nan in NP

[Antigr]

There are warnings about divide by zero errors I modified the code a bit so as to find the image file on my computer, but nothing else:

 import numpy as np from PIL import Image import math import os #np.seterr(divide='ignore', invalid='ignore') path = os.path.dirname(__file__) os.chdir(path) path = os.getcwd() image = Image.open(os.path.abspath('../images/test.jpg')) # image=Image.open('E:/test.jpg') imgr,imgg,imgb=image.split() r=np.asarray(imgr).astype('float64') g=np.asarray(imgg).astype('float64') b=np.asarray(imgb).astype('float64') vari=(g-r)/(g+r-b) nan=0 a=[] for i in range (len(vari)): for j in range(len(vari[i])): if not np.isnan(vari[i][j]): b=float(vari[i][j]) a.append(b) else: vari[i][j]=1 nan=nan+1 print('len vari',len(vari)*len(vari[0])) print('len a (no NaN elements)',len(a)) print('TOTAL NAN',nan) for i in range (len(vari)): for j in range(len(vari[i])): if np.isnan(vari[i][j]): print ('new NAN object',i,j) s=0 vari=vari.ravel() for i in range (len(a)): if math.isnan(a[i]): print ('new NAN object in a massiv',i,j) for i in range(len(a)): s=s+a[i] print ('summa iz a',s) s2=0 for i in range(len(vari)): s2=s2+vari[i] print ('summa',s2) 

run:

Output:
 Microsoft Windows [Version 6.1.7601] Copyright (c) 2009 Microsoft Corporation. All rights reserved. m:/python/e-h/f/forum/g-m/ImageSoftware/src/June26-2018.py:18: RuntimeWarning: divide by zero encountered in true_divide vari=(g-r)/(g+r-b) m:/python/e-h/f/forum/g-m/ImageSoftware/src/June26-2018.py:18: RuntimeWarning: invalid value encountered in true_divide vari=(g-r)/(g+r-b) len vari 2666000 len a (no NaN elements) 2665987 TOTAL NAN 13 summa iz a nan m:/python/e-h/f/forum/g-m/ImageSoftware/src/June26-2018.py:46: RuntimeWarning: invalid value encountered in double_scalars s2=s2+vari[i] summa nan 

yes i know about it, but in a array there is no NaN elements, and sum is 0