Median and uniform filters in SciPy - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of median filter on 1D array

What is the output of the following code applying a median filter with size 3 on a 1D array?

SciPy

import numpy as np
from scipy.ndimage import median_filter

arr = np.array([1, 2, 80, 4, 5, 6, 7])
filtered = median_filter(arr, size=3)
print(filtered.tolist())

A [2, 2, 4, 5, 6, 6, 7]

B [1, 2, 80, 4, 5, 6, 7]

C [1, 2, 4, 4, 5, 6, 7]

D [1, 2, 4, 5, 5, 6, 7]

💡 Hint

Median filter replaces each element with the median of neighbors within the window size.

The median filter with size 3 looks at each element and its neighbors (previous and next). For example, at index 2, neighbors are [2, 80, 4], median is 4. This smooths out the spike at 80.

❓ data_output

intermediate

Result of uniform filter on 2D array

What is the resulting 2D array after applying a uniform filter with size 3 on this 3x3 array?

SciPy

import numpy as np
from scipy.ndimage import uniform_filter

arr = np.array([[1, 2, 3],
                [4, 5, 6],
                [7, 8, 9]])
filtered = uniform_filter(arr, size=3, mode='constant', cval=0)
print(filtered)

A

[[1.55555556 2.33333333 1.88888889]
 [3.66666667 5.         4.11111111]
 [3.11111111 4.66666667 3.55555556]]

B

[[1.33333333 2.         1.66666667]
 [3.         5.         3.        ]
 [2.66666667 4.         2.66666667]]

C

[[1. 2. 3.]
 [4. 5. 6.]
 [7. 8. 9.]]

D

[[0.66666667 1.33333333 1.        ]
 [2.33333333 4.         3.        ]
 [2.         3.33333333 2.33333333]]

💡 Hint

Uniform filter computes the average over the window including zero padding outside edges.

The uniform filter with size 3 averages each 3x3 neighborhood. Since mode='constant' with cval=0, zeros pad the edges, lowering averages near borders.

❓ visualization

advanced

Visual effect of median vs uniform filter on noisy image

Which option best describes the visual difference after applying median and uniform filters on a noisy grayscale image?

SciPy

import numpy as np
import matplotlib.pyplot as plt
from scipy.ndimage import median_filter, uniform_filter

np.random.seed(0)
image = np.ones((10,10)) * 100
noise = np.random.randint(0, 50, (10,10))
noisy_image = image + noise
median_img = median_filter(noisy_image, size=3)
uniform_img = uniform_filter(noisy_image, size=3)

plt.subplot(1,3,1)
plt.title('Noisy')
plt.imshow(noisy_image, cmap='gray')
plt.subplot(1,3,2)
plt.title('Median Filter')
plt.imshow(median_img, cmap='gray')
plt.subplot(1,3,3)
plt.title('Uniform Filter')
plt.imshow(uniform_img, cmap='gray')
plt.show()

A Median filter increases noise; uniform filter sharpens edges.

B Median filter removes noise while preserving edges; uniform filter blurs edges and smooths noise.

C Both filters produce identical smoothing and edge preservation.

D Uniform filter removes noise while preserving edges; median filter blurs edges and smooths noise.

💡 Hint

Median filter is good for salt-and-pepper noise; uniform filter averages values.

Median filter replaces each pixel with median of neighbors, preserving edges better. Uniform filter averages neighbors, causing blur and edge smoothing.

🧠 Conceptual

advanced

Effect of filter size on median and uniform filters

How does increasing the filter size affect the output of median and uniform filters on a noisy signal?

A Increasing size increases smoothing for both filters; median filter preserves edges better but may remove small details.

B Increasing size decreases smoothing for both filters; median filter blurs edges more than uniform filter.

C Increasing size has no effect on median filter; uniform filter smoothing decreases.

D Increasing size causes median filter to amplify noise; uniform filter preserves edges perfectly.

💡 Hint

Think about how larger windows affect averaging and median calculations.

Larger filter sizes mean more neighbors considered, increasing smoothing. Median filter still preserves edges better but may remove small features as size grows.

🔧 Debug

expert

Identify the error in applying median filter on a 3D array

What error will this code raise when applying median_filter with size=3 on a 3D array without specifying axis?

SciPy

import numpy as np
from scipy.ndimage import median_filter

arr = np.random.randint(0, 10, (4,4,4))
filtered = median_filter(arr, size=3)
print(filtered.shape)

A ValueError: size must be less than or equal to input size along each axis

B TypeError: size must be a sequence of integers

C No error; output shape is (4,4,4)

D No error; output shape is (2,2,2)

💡 Hint

Check how median_filter handles size parameter for multi-dimensional arrays.

median_filter accepts an integer size and applies it to all axes. The output shape matches input shape. No error occurs.