How to slice region of numpy array surrounding minimum value (i.e. 100 pixels on each side of a point)

Question

I have a 2D numpy array which is image data, and contains a dot with a lower pixel value than the rest of the image. This dot is not necessarily in the centre of the image. I've found the coordinates of the dot from my dataset ('data', Array of uint16, 1024x1024) using:

centre = unravel_index(data.argmin(), data.shape)

What I want to do now is keep 100 pixels on each side of the dot (top, bottom, left and right) but crop out the rest of the image. This region has a different background value which I need to keep, so I don't want to just keep the dot and fill it with zeros, I want to keep the existing data in the array in this region.

I've been searching for answers for a few hours but can't find any that really match what I'm trying to do. All the solutions I've found seem to replace all values in the array with the minimum value.

Answer 1

Once you found the indices of your centre pixel, it is simple to get the indices of the bounding box of the new region around those indices. A straightforward way is the following (example case with a smaller shape than yours for illustration purposes):

import numpy as np

# Example data
nx, ny = 30, 15   # Dimensions of data array
width  = 3        # Number of extra pixels in each direction
np.random.seed(1)
data = np.random.random((ny,nx))    

# Find centre indices
centre = np.unravel_index(data.argmin(), data.shape)  
# Find indices of new borders
idx_left  = max(centre[1] - width, 0)
idx_right = min(centre[1] + width + 1, nx - 1)
if idx_right == nx - 1:  # When the last index in x-dimension is selected
    idx_right = None
idx_bot   = max(centre[0] - width, 0) 
idx_top   = min(centre[0] + width + 1, ny - 1)
if idx_top == ny - 1:    # When the last index in y-dimension is selected
    idx_top = None
# Output
view = data[idx_bot:idx_top, idx_left:idx_right]

Note however that there may be functions that already do this (better) in image processing libraries (I do not have experience with them though).