Image sharpening in Signal Processing

Introduction

Image sharpening makes pictures clearer by highlighting edges and details. It helps us see important parts better.

When a photo looks blurry and you want to make it clearer.

To enhance details in medical images for better diagnosis.

Improving text readability in scanned documents.

Making features stand out in satellite images.

Before applying other image analysis to get better results.

Syntax

Signal Processing

sharpened_image = original_image + alpha * (original_image - blurred_image)

This formula is called unsharp masking.

Alpha controls how strong the sharpening effect is.

Examples

This example sharpens a grayscale photo using Gaussian blur and adds details back.

Signal Processing

import cv2
import numpy as np

original = cv2.imread('photo.jpg', cv2.IMREAD_GRAYSCALE)
blurred = cv2.GaussianBlur(original, (5,5), 0)
alpha = 1.5
sharpened = cv2.addWeighted(original, 1 + alpha, blurred, -alpha, 0)

Using SciPy to sharpen a numpy image array with Gaussian smoothing.

Signal Processing

from scipy.ndimage import gaussian_filter

original = image_array
blurred = gaussian_filter(original, sigma=1)
alpha = 1.0
sharpened = original + alpha * (original - blurred)

Sample Program

This code creates a small image, blurs it, and sharpens it using unsharp masking. It prints all steps so you can see the effect.

Signal Processing

import cv2
import numpy as np

# Create a simple 5x5 grayscale image with a bright square in the center
original = np.array([
  [50, 50, 50, 50, 50],
  [50, 200, 200, 200, 50],
  [50, 200, 255, 200, 50],
  [50, 200, 200, 200, 50],
  [50, 50, 50, 50, 50]
], dtype=np.uint8)

# Blur the image using Gaussian blur
blurred = cv2.GaussianBlur(original, (3,3), 0)

# Sharpening factor
alpha = 1.0

# Apply unsharp masking
sharpened = cv2.addWeighted(original, 1 + alpha, blurred, -alpha, 0)

print('Original Image:\n', original)
print('\nBlurred Image:\n', blurred)
print('\nSharpened Image:\n', sharpened)

Important Notes

Too much sharpening can make images look unnatural or noisy.

Sharpening works best on images that are slightly blurry.

Use grayscale images first to understand sharpening before color images.

Summary

Image sharpening highlights edges to make pictures clearer.

Unsharp masking is a common way to sharpen images by subtracting a blurred version.

Adjust the sharpening strength carefully to avoid unwanted noise.

Practice

1. What is the main purpose of image sharpening in signal processing?

easy

A. To reduce the image size

B. To highlight edges and make the image clearer

C. To blur the image for artistic effect

D. To convert the image to grayscale

Solution

1. Step 1: Understand image sharpening

   Image sharpening enhances edges and details to make images clearer.

2. Step 2: Compare options

   Only To highlight edges and make the image clearer describes highlighting edges and clarity, which is the goal of sharpening.

3. Final Answer:

   To highlight edges and make the image clearer -> Option B

4. Quick Check:

   Image sharpening = highlight edges [OK]

Hint: Sharpening means making edges clearer, not blurring [OK]

Common Mistakes:

• Confusing sharpening with blurring
• Thinking sharpening changes image size
• Assuming sharpening converts color formats

2. Which of the following is the correct Python syntax to apply a simple sharpening kernel using convolution with OpenCV?

easy

A. cv2.sharpen(image, kernel)

B. cv2.convolve(image, kernel)

C. cv2.filter2D(image, -1, kernel)

D. cv2.applySharpen(image, kernel)

Solution

1. Step 1: Recall OpenCV function for convolution

   OpenCV uses cv2.filter2D() to apply kernels for filtering images.

2. Step 2: Check options

   Only cv2.filter2D(image, -1, kernel) uses cv2.filter2D with correct parameters; others are invalid functions.

3. Final Answer:

   cv2.filter2D(image, -1, kernel) -> Option C

4. Quick Check:

   OpenCV convolution = filter2D [OK]

Hint: Use cv2.filter2D for custom kernel convolution [OK]

Common Mistakes:

• Using non-existent OpenCV functions
• Confusing convolution with sharpening function names
• Incorrect parameter order

3. Given the following Python code using OpenCV, what will be the output type of sharpened?

import cv2
import numpy as np
image = cv2.imread('photo.jpg')
kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]])
sharpened = cv2.filter2D(image, -1, kernel)

medium

A. A sharpened image as a NumPy array

B. An error due to kernel shape

C. A grayscale image

D. A blurred image

Solution

1. Step 1: Understand filter2D output

   cv2.filter2D returns an image as a NumPy array after applying the kernel.

2. Step 2: Analyze kernel and output

   The kernel is a sharpening kernel, so output is sharpened image, still in color and array form.

3. Final Answer:

   A sharpened image as a NumPy array -> Option A

4. Quick Check:

   filter2D output = NumPy array image [OK]

Hint: filter2D returns image array, not blurred or grayscale [OK]

Common Mistakes:

• Assuming output is blurred instead of sharpened
• Thinking output is grayscale without conversion
• Believing kernel shape causes error

4. Identify the error in this Python code snippet for image sharpening:

import cv2
import numpy as np
image = cv2.imread('img.png')
kernel = np.array([[0, -1, 0], [-1, 5, -1], [0, -1, 0]])
sharpened = cv2.filter2D(image, 1, kernel)
cv2.imshow('Sharpened', sharpened)
cv2.waitKey(0)
cv2.destroyAllWindows()

medium

A. The ddepth parameter should be -1, not 1

B. The kernel shape is invalid

C. cv2.imshow cannot display color images

D. cv2.waitKey is missing argument

Solution

1. Step 1: Check ddepth parameter in filter2D

   ddepth=-1 means output image has same depth as input; 1 is invalid and may cause errors.

2. Step 2: Validate other code parts

   Kernel shape is valid, cv2.imshow supports color, waitKey(0) is correct usage.

3. Final Answer:

   The ddepth parameter should be -1, not 1 -> Option A

4. Quick Check:

   ddepth=-1 for same depth [OK]

Hint: Use ddepth=-1 in filter2D to keep image depth [OK]

Common Mistakes:

• Using wrong ddepth value causing errors
• Thinking kernel shape is wrong
• Misunderstanding cv2.imshow capabilities

5. You want to sharpen an image using unsharp masking. Which sequence of steps correctly applies this method?

hard

A. Directly apply a sharpening kernel without blurring

B. Add noise to the image, then apply a sharpening kernel

C. Convert image to grayscale, then apply a sharpening kernel

D. Blur the image, subtract the blurred from original, then add result back to original

Solution

1. Step 1: Understand unsharp masking process

   Unsharp masking involves blurring the image, subtracting this blur from original to get edges, then adding edges back.

2. Step 2: Evaluate options

   Only Blur the image, subtract the blurred from original, then add result back to original correctly describes this sequence; others miss key steps or add unrelated actions.

3. Final Answer:

   Blur the image, subtract the blurred from original, then add result back to original -> Option D

4. Quick Check:

   Unsharp masking = blur, subtract, add [OK]

Hint: Unsharp masking = blur, subtract, add back edges [OK]

Common Mistakes:

• Skipping the blur step
• Adding noise instead of subtracting blur
• Applying sharpening kernel directly without unsharp steps