0

I am working with various images and I wanted to implement a method but I have no idea where to start. So I am looking forward to some suggestions.

My idea as a method is to take an image array and after selecting a region of interest and then we say the selectedROI is 10x10 region of pixels as a array. Then how can I make a filter of size say for example 2x5 and after making the 2x5 filter how to move the filter throughout the selectedROI and comparing each 2x5 filters for min, max & mean values.

I have successfully achieved the selecting the ROI and printing the selecedROI as a array. All I want to do in simple words is to select a bunch of pixels and compare the bunch with another bunch within the whole selectedROI. How is it possible.

Blueman0002
  • 39
  • 7

1 Answers1

2

It sounds like you have an array of pixels with shape (10, 10), and want to slide a window of shape (2, 5) across the array, performing an operation at each position.

NumPy has operations to help with this (see here: https://numpy.org/devdocs/reference/generated/numpy.lib.stride_tricks.sliding_window_view.html), however the documentation says sliding_window_view is generic and thus nonoptimal for your case. They suggest using scipy.ndimage operations instead, which I have shown examples of below. https://docs.scipy.org/doc/scipy/reference/reference/generated/scipy.ndimage.generic_filter.html

To start, I have imported the packages and defined a 10x10 array and 2x5 filter size.

>>> import numpy as np
>>> from scipy.ndimage import generic_filter
>>> selectedROI = np.array(range(100)).reshape((10, 10))
>>> selectedROI
array([[ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
       [20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
       [30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
       [40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
       [50, 51, 52, 53, 54, 55, 56, 57, 58, 59],
       [60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
       [70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
       [80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
       [90, 91, 92, 93, 94, 95, 96, 97, 98, 99]])
>>> filter_size = (2, 5)

Any function can be applied in sliding-window fashion with generic_filter (np.average used here). Variables a,b,c,d were chosen because generic_filter produces values for all positions in your input array, including positions where the window does not fall entirely within the input array. For an explanation of the "or None" in the slicing, see here: numpy negative indexing a[:-0]

>>> a = filter_size[0]//2
>>> b = (1 - filter_size[0]%2) - (filter_size[0]//2)
>>> c = filter_size[1]//2
>>> d = (1 - filter_size[1]%2) - (filter_size[1]//2)
>>> generic_filter(selectedROI, np.average, filter_size)[a:b or None, c:d or None]
array([[ 7,  8,  9, 10, 11, 12],
       [17, 18, 19, 20, 21, 22],
       [27, 28, 29, 30, 31, 32],
       [37, 38, 39, 40, 41, 42],
       [47, 48, 49, 50, 51, 52],
       [57, 58, 59, 60, 61, 62],
       [67, 68, 69, 70, 71, 72],
       [77, 78, 79, 80, 81, 82],
       [87, 88, 89, 90, 91, 92]])

You can use other functions, such as np.min and np.max for your use case, or write your own custom functions as shown here: How do I use scipy.ndimage.filters.gereric_filter?

>>> generic_filter(selectedROI, np.min, filter_size)[a:b or None, c:d or None]
array([[ 0,  1,  2,  3,  4,  5],
       [10, 11, 12, 13, 14, 15],
       [20, 21, 22, 23, 24, 25],
       [30, 31, 32, 33, 34, 35],
       [40, 41, 42, 43, 44, 45],
       [50, 51, 52, 53, 54, 55],
       [60, 61, 62, 63, 64, 65],
       [70, 71, 72, 73, 74, 75],
       [80, 81, 82, 83, 84, 85]])
>>> generic_filter(selectedROI, np.max, filter_size)[a:b or None, c:d or None]
array([[14, 15, 16, 17, 18, 19],
       [24, 25, 26, 27, 28, 29],
       [34, 35, 36, 37, 38, 39],
       [44, 45, 46, 47, 48, 49],
       [54, 55, 56, 57, 58, 59],
       [64, 65, 66, 67, 68, 69],
       [74, 75, 76, 77, 78, 79],
       [84, 85, 86, 87, 88, 89],
       [94, 95, 96, 97, 98, 99]])

These examples used 2D arrays for clarity, but the operations should extend to 3D pixel arrays.

R B
  • 61
  • 4
  • Hello, Thank you so much this is very insightful on where can I start with. Thank you so much. – Blueman0002 Aug 12 '21 at 07:55
  • I think I am extending my luck here. But as far as I understand theoretically can anyone use this method in a edge detected image and perform filters within the detected edges right? – Blueman0002 Aug 12 '21 at 08:01
  • Yes, you can replace np.average with any custom function, which you can write to perform filters in an edge detected image. I would suggest checking out this link to see prewritten functions for many common use cases, if you want to avoid writing the custom function yourself. https://docs.scipy.org/doc/scipy/reference/reference/ndimage.html#module-scipy.ndimage – R B Aug 13 '21 at 12:21
  • Ohhh. That is interesting. I only need average function . Man than you so much you saved my day with these. Have a great weekend and take care. – Blueman0002 Aug 14 '21 at 15:37