CODEWITHSUNDEEP
pythonimagenumpyscipynormalization
melatonin15
melatonin15

Reputation: 2269

Numpy ndarray image pixel mean for pixel values greater than zero: Normalizing image

I am trying to read and normalize a 3 channel image in numpy. For each channel in the image, I want to calculate mean of the pixel values which are greater than zero.

I started with:

from scipy import misc
img = misc.imread('test.png')
print(type(img) ) #<type 'numpy.ndarray'>
print(img.shape) #(512, 512, 3)

But I am not sure first 1.) how to index out positive values preserving dimension and without flattening the array. And 2.) How to take channel wise mean of the selected positive values.

My full normalization process is like:

img_mean = mean(img[img >0])#channel wise mean of positive pixels
img_std  = std(img[img>0]) #channel wise std. deviation of positive pixels
img_norm = (img - img_mean)/img_std
img_norm[img_norm < -1] = 0 #setting pixel values less than 1 to 0.

Here is an example of image I am working with

enter image description here

Upvotes: 1

Views: 1766

Answers (1)

Divakar
Divakar

Reputation: 221574

Easiest way would be to mask out all zeros as NaNs and then use np.nanmean and np.nanstd to essentially ignore the zeros from the calculations, like so -

imgn = np.where(img>0,img,np.nan)
img_norm = (img - np.nanmean(imgn,axis=(0,1)))/np.nanstd(imgn,axis=(0,1))

Upvotes: 2

Related Questions