Min filter in Python using OpenCV

By min filter I guess you mean running a kernel through each location in an image and replacing the kernel centre with the min value within the kernel's pixels.

To achieve this in Opencv you simply can use, cv2.erode. Documentation here.

But first, you need to define the size and shape of the kernel. You use cv.getStructuringElement doc here:

Example:

size = (3, 3)
shape = cv2.MORPH_RECT
kernel = cv2.getStructuringElement(shape, size)
min_image = cv2.erode(image, kernel)

The morphological erosion is a minimum filter. In OpenCV this is implemented in the function erode.

In addition to the previous answers, I implement in python + opencv the code that applies the minimum and maximum box filter.

import cv2

def minimumBoxFilter(n, path_to_image):
  img = cv2.imread(path_to_image)

  # Creates the shape of the kernel
  size = (n, n)
  shape = cv2.MORPH_RECT
  kernel = cv2.getStructuringElement(shape, size)

  # Applies the minimum filter with kernel NxN
  imgResult = cv2.erode(img, kernel)

  # Shows the result
  cv2.namedWindow('Result with n ' + str(n), cv2.WINDOW_NORMAL) # Adjust the window length
  cv2.imshow('Result with n ' + str(n), imgResult)


def maximumBoxFilter(n, path_to_image):
  img = cv2.imread(path_to_image)

  # Creates the shape of the kernel
  size = (n,n)
  shape = cv2.MORPH_RECT
  kernel = cv2.getStructuringElement(shape, size)

  # Applies the maximum filter with kernel NxN
  imgResult = cv2.dilate(img, kernel)

  # Shows the result
  cv2.namedWindow('Result with n ' + str(n), cv2.WINDOW_NORMAL) # Adjust the window length
  cv2.imshow('Result with n ' + str(n), imgResult)


if __name__ == "__main__":
  path_to_image = 'images/africa.tif'

  print("Test the function minimumBoxFilter()")
  minimumBoxFilter(3, path_to_image)
  minimumBoxFilter(5, path_to_image)
  minimumBoxFilter(7, path_to_image)
  minimumBoxFilter(11, path_to_image)
  cv2.waitKey(0)
  cv2.destroyAllWindows()

  print("Test the function maximumBoxFilter()")
  maximumBoxFilter(3, path_to_image)
  maximumBoxFilter(5, path_to_image)
  maximumBoxFilter(7, path_to_image)
  maximumBoxFilter(11, path_to_image)
  cv2.waitKey(0)
  cv2.destroyAllWindows()

where path_to_image is a string. For example: 'images/africa.tif'

I obtain the following results:

With the minimum filter:

With the maximum filter: