I am trying to break down an image into smaller images (like a grid), process those blocks and join them back up for display, using the reshape function.
I am currently doing this by means of a nested for loop as shown in the code. However, as built-in functions like reshape()/np.reshape() are faster compared to for-loops, I want to achieve the same result by using the reshape() function. I am dealing with a 3-channel BGR image and would like several smaller 3-channel BGR images, that are pieces of the grid that splits the big one.
import numpy as np
roi_ht = 2
roi_wd = 2
parent_array = np.array([[1,10,2,20],
[100,1000,200,2000],
[3,30,4,40],
[300,3000,400,4000]])
smaller_images = [parent_array[x*roi_ht:(x+1)*roi_ht, y*roi_wd:(y+1)*roi_wd] for x in range(0,2) for y in range(0,2)]
This is the desired result to be achieved using 'reshape' or any other function built-in to python/numpy (smaller images) :
[array([[ 1, 10], [ 100, 1000]]),
array([[ 2, 20], [ 200, 2000]]),
array([[ 3, 30], [ 300, 3000]]),
array([[ 4, 40], [ 400, 4000]])]
