What is Image Segmentation?

Image segmentation is the process of partitioning an image into multiple segments or regions, which makes it easier to analyze and interpret. By separating an image into distinct parts, we can isolate objects of interest, enhance image analysis, and improve recognition systems. It's a fundamental step in various applications, including object detection, image editing, and medical imaging.

Why is Image Segmentation Important?

• Object Recognition: Segmentation helps in identifying and locating objects within an image, which is vital for tasks like facial recognition and autonomous driving.
• Medical Imaging: In healthcare, image segmentation aids in isolating specific areas in MRI or CT scans, allowing for better diagnosis and treatment planning.
• Image Processing: It simplifies complex images, making it easier to apply filters, enhancements, or modifications.

Techniques for Image Segmentation

There are various techniques for image segmentation, but in this blog, we will focus on the following three widely-used methods:

1. Thresholding
2. K-means Clustering
3. Watershed Algorithm

1. Thresholding

Thresholding is one of the simplest methods of image segmentation. It converts a grayscale image into a binary image by setting a threshold value. All pixel values above the threshold become white, and all pixel values below it become black.

Example of Thresholding

import cv2
import numpy as np

# Load image in grayscale
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)

# Apply thresholding
_, thresholded = cv2.threshold(image, 127, 255, cv2.THRESH_BINARY)

# Display the results
cv2.imshow('Original Image', image)
cv2.imshow('Thresholded Image', thresholded)
cv2.waitKey(0)
cv2.destroyAllWindows()

In the above example, we use OpenCV to load an image, apply a simple threshold, and then display both the original and thresholded images.

2. K-means Clustering

K-means clustering is a popular unsupervised learning algorithm that can be used for image segmentation. It works by clustering pixel values into K different groups, which can highlight distinct segments in the image.

Example of K-means Clustering

import cv2

# Load image
image = cv2.imread('image.jpg')
pixel_values = image.reshape((-1, 3))
pixel_values = np.float32(pixel_values)

# Define K-means criteria and apply k-means
k = 3
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 100, 0.2)
_, labels, centers = cv2.kmeans(pixel_values, k, None, criteria, 10, cv2.KMEANS_RANDOM_CENTERS)

# Convert back to uint8
centers = np.uint8(centers)
segmented_image = centers[labels.flatten()]
segmented_image = segmented_image.reshape(image.shape)

# Display the results
cv2.imshow('Original Image', image)
cv2.imshow('Segmented Image', segmented_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

In this example, we reshape the image into a 2D array where each pixel's color is represented as a 3D point (R, G, B). We then apply K-means to categorize pixels into three clusters, effectively segmenting the image.

3. Watershed Algorithm

The watershed algorithm is a powerful image segmentation technique that treats the grayscale image as a topographic surface. The algorithm identifies "catchment basins" based on gradients in pixel intensity and segments objects accordingly.

Example of Watershed Algorithm

import cv2
import numpy as np

# Load image in grayscale
image = cv2.imread('image.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

# Apply a binary threshold for segmentation
_, thresh = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY_INV)

# Find sure foreground area
dist_transform = cv2.distanceTransform(thresh, cv2.DIST_L2, 5)
_, sure_fg = cv2.threshold(dist_transform, 0.7 * dist_transform.max(), 255, 0)

# Find unknown region
unknown = cv2.subtract(thresh, np.uint8(sure_fg))

# Marker labelling
_, markers = cv2.connectedComponents(np.uint8(sure_fg))

# Add one to all the labels so that the sure regions are marked with 1
markers = markers + 1
markers[unknown == 255] = 0

# Mark the unknown regions with zero

# Apply the watershed algorithm
markers = cv2.watershed(image, markers)
image[markers == -1] = [255, 0, 0]

# Mark boundaries with red

# Display results
cv2.imshow('Original Image', image)
cv2.imshow('Watershed Segmentation', markers)
cv2.waitKey(0)
cv2.destroyAllWindows()

In this example, we perform segmentation using the watershed algorithm by creating markers and highlighting the boundaries of different segments.

Conclusion

Image segmentation is a powerful technique that underpins many computer vision applications. With tools like OpenCV in Python, it's easier than ever to implement various segmentation methods to suit your needs. Each technique has its pros and cons, depending on the complexity of the images and the desired application. By experimenting with these methods, you can gain valuable insight into how computer vision works, paving the way for more advanced projects.