When you read the image using cv2.imread(), it will return an array of numbers. This array denotes the image and numbers denote the values in each pixel. For a one-channel image (black&white ), one number denotes each pixel 2D, but for an three-channel image (colored), three images denote each pixel 3D.

Hence the indices of the array give you the location of each pixel. Given that each block is 10 X 10 pixel each than (0,0) will be image[0:10,0:10] and (0,1) will image[0:10,10:20]. For a box at x,y with size S, pixels will be image[x*S : (x+1)*S , y*S : (y+1)*S].

In the article I’ve used the code :

block = img[h0:h,w0:w]

Where :

h = CELL_SIZE*(coordinate[0]+1)
w = CELL_SIZE*(coordinate[1]+1)
h0= CELL_SIZE*coordinate[0]
w0= CELL_SIZE*coordinate[1]

Hope, your problem is solved.