Morphological Operations with OpenCV

Install them with:

pip install opencv-python matplotlib numpy

Basic Setup

import cv2
import numpy as np
import matplotlib.pyplot as plt

Creating a Base Image

First, let's create a black image with white text to demonstrate morphological operations:

def createImage():
    img = np.zeros(shape=(600,600))
    font = cv2.FONT_HERSHEY_COMPLEX
    cv2.putText(img, text='OMER', org=(50,300), 
                fontFace=font, fontScale=5, 
                color=(255,255,255), thickness=25, 
                lineType=cv2.LINE_AA)
    return img

def show(img):
    fig = plt.figure(figsize=(12,10))
    ax = fig.add_subplot(111)
    ax.imshow(img, cmap='gray')
    plt.show()

image = createImage()
show(image)

Morphological Operations

1. Erosion

Erosion is a morphological operation used in image processing to shrink the white (foreground) regions in a binary image. It works by sliding a small matrix called a kernel (usually 3x3) over the image. At each position, if all the pixels under the kernel are white (255), the center pixel remains white; otherwise, it is set to black (0). This causes the boundaries of objects to erode or shrink, removing small white noise and thinning the shapes. Erosion is commonly used to clean up binary images, especially before applying contour detection or other analysis.

kernel = np.ones(shape=(5,5), dtype=np.uint8)
result = cv2.erode(image, kernel, iterations=5)
show(result)

2. Handling White Noise

Let's add and remove white noise (salt noise):

The code demonstrates how to add and remove white noise (specifically, "salt" noise) from a binary image using morphological operations in OpenCV:

1. Adding White Noise:

   • white_noise = np.random.randint(low=0, high=2, size=(600,600))
     This creates a 600x600 array of random 0s and 1s.
   • white_noise = white_noise * 255
     Converts the 1s to 255, so the noise is either black (0) or white (255).
   • noise_image = white_noise + image
     Adds the white noise to the original image, introducing random white pixels ("salt" noise).

2. Removing Noise with Opening:

   • opening = cv2.morphologyEx(noise_image, cv2.MORPH_OPEN, kernel)
     Applies the "opening" morphological operation, which is an erosion followed by a dilation. This operation removes small white noise dots while preserving the main structure of the text.

Note:

What does MORPH_OPEN do?

MORPH_OPEN is a morphological operation in OpenCV that performs an erosion followed by a dilation using the same kernel. Its main purpose is to remove small white noise (tiny white dots or regions) from a binary image while preserving the overall shape and size of larger white objects.

• Erosion removes small white regions (noise) by shrinking the white areas.
• Dilation then expands the remaining white regions back to their original size, but without the small noise that was removed.

This makes MORPH_OPEN especially useful for cleaning up images before further processing, such as contour detection or OCR.

In summary:
cv2.MORPH_OPEN helps to clean up small white noise in binary images by eroding and then dilating, preserving the main shapes but removing tiny artifacts.

Image source: docs.opencv.org

Summary:
This code shows how to simulate salt noise on an image and then clean it up using the morphological "opening" operation, which is useful for preprocessing images before further analysis.

3. Handling Black Noise

Now let's add and remove black noise

black_noise = np.random.randint(low=0, high=2, size=(600,600))
black_noise = black_noise * -255
black_noise_image = black_noise + image
black_noise_image[black_noise_image == -255] = 0

# Remove noise using closing (dilation followed by erosion)
closing = cv2.morphologyEx(black_noise_image, cv2.MORPH_CLOSE, kernel)
show(closing)

4. Morphological Gradient

The gradient shows the boundaries of the text: