a non-vectorized and semi-vectorized implementation for calculating std for a batch of images. the semi-vectorized is the one, one should use, since its as fast as the numpy.std

calculate_std.py

#In the name of God, the most compassionate the most merciful
#a non-vectorized and semi-vectorized implementation for calculating std for a batch of images.
#the semi-vectorized is the one, one should use, since its as fast as the numpy.std  
#Seyyed Hossein Hasan pour 
#Coderx7@gmail.com
#1/12/2017
import math
#unvectorized version --really slow!
def calc_std_classic(a):
    #sum all elements in each channel and divide by the number of elements
    batch = a.shape[0]
    channel = a.shape[1]
    width = a.shape[2]
    height = a.shape[3]
    #this method is uploaded here : https://gist.github.com/Coderx7/de210643b15e206a4fd6a5da9f7c4d2b
    mean = calc_mean_classic2(a)
 #https://www.daniweb.com/programming/software-development/threads/348002/mean-and-standard-deviation-of-image
 #https://docs.scipy.org/doc/numpy/reference/generated/numpy.std.html   
    sum = np.zeros((channel))
    for i in range(batch):
        for j in range(channel):
            for w in range(width):
                for h in range(height):
                    sum[j] += (abs(a[i,j,w,h] - mean[j])**2)
                    
    var = (sum/(width*height*batch))
    std = [round(math.sqrt(x),8) for x in var ]
    return std
  
#semi-vectorized, very fast. use this if you face memory shortage errors because of your dataset being too big for your memory
def calc_std_classic2(a):
    batch = a.shape[0]
    channel = a.shape[1]
    width = a.shape[2]
    height = a.shape[3]
    mean = calc_mean_classic2(a)
    
    sum = np.zeros((channel))
    for i in range(batch):
        for j in range(channel):
            sum[j] += np.sum(abs(a[i,j,:,:] - mean[j])**2)
                    
    var = (sum/(width*height*batch))
    std = [round(math.sqrt(x),8) for x in var ]
    
    return std