How to perform Otsu’s thresholding on an image using Python OpenCV?

Image thresholding is a widely used method in image analysis and computer vision applications. This technique is used to segment an image into different parts or regions based on the intensity of the pixels. One of the most popular methods for image thresholding is Otsu’s method, which tries to find a threshold value that can separate the foreground from the background in an image. In this article, we will discuss how to perform Otsu’s thresholding on an image using Python OpenCV.

What is Otsu’s Thresholding?

Otsu’s thresholding is a technique that aims to find the threshold value that minimizes the variance of the foreground and background pixels. This method assumes that the image can be separated into two classes, foreground and background, based on a threshold value. The threshold value is chosen in such a way that the variance between the foreground and background pixels is minimized.

The Otsu’s thresholding algorithm takes the histogram of the image as input, which is a distribution of the pixel intensities in the image. The histogram is then normalized to compute the probability distribution of the pixel values. The algorithm then computes the cumulative sums of the probabilities and the means of the pixel intensities for the foreground and background classes. Finally, the algorithm computes the between-class variance for each possible threshold value and chooses the threshold that minimizes this value.

Implementing Otsu’s Thresholding in Python OpenCV

To demonstrate how to perform Otsu’s thresholding on an image using Python OpenCV, we will start by importing the necessary libraries and reading an image.

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

# Read the image
img = cv2.imread('image.png', 0)

In the above code, we have imported the required libraries, including OpenCV, numpy, and matplotlib. We have also read an image using the cv2.imread() function. The second argument of this function specifies the color mode of the image, which is set to 0 to read the image in grayscale mode.

Next, we will compute the histogram of the image using the cv2.calcHist() function.

# Compute the histogram of the image
hist = cv2.calcHist([img], [0], None, [256], [0, 256])

The cv2.calcHist() function takes the following arguments:

• images: The input image in the form of a list.
• channels: The channel(s) of the image for which the histogram needs to be computed. For grayscale images, this value is [0].
• mask: The mask to be applied to the image. If None, no mask will be applied.
• histSize: The number of bins in the histogram.
• ranges: The minimum and maximum values to be considered for the histogram.

In the above code, we have computed the histogram of the image for 256 bins ranging from 0 to 256.

Now that we have computed the histogram, we can plot it using the matplotlib library.

# Plot the histogram of the image
plt.hist(img.ravel(), 256, [0, 256])
plt.show()

The plt.hist() function takes the following arguments:

• x: The values to be considered for the histogram. In our case, we have used img.ravel(), which flattens the image into a 1D array.
• bins: The number of bins in the histogram.
• range: The minimum and maximum values to be considered for the histogram.

Next, we will implement Otsu’s thresholding algorithm to find the optimal threshold value.

# Find the Otsu's threshold value
ret, thresh = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

The cv2.threshold() function takes the following arguments:

• src: The input image.
• thresh: The threshold value to be used.
• maxval: The maximum value to be used if the pixel value is greater than the threshold value.
• type: The thresholding technique to be used. In our case, we have used cv2.THRESH_BINARY + cv2.THRESH_OTSU, which performs binary thresholding using Otsu’s method.

The above code will compute the thresholded image using the optimal threshold value obtained from Otsu’s method.

Finally, we will display the input and thresholded images using the matplotlib library.

# Display# the input and thresholded images
fig, axs = plt.subplots(1, 2, figsize=(10, 10))
axs[0].imshow(img, cmap='gray')
axs[0].set_title('Input Image')
axs[0].axis('off')
axs[1].imshow(thresh, cmap='gray')
axs[1].set_title('Thresholded Image')
axs[1].axis('off')
plt.show()

The above code will display both the input and thresholded images side by side, allowing us to compare the differences.

Conclusion

In this article, we discussed how to perform Otsu’s thresholding on an image using Python OpenCV. We explained the concept of Otsu’s thresholding and how it works by finding the threshold value that minimizes the variance of the foreground and background pixels. We also provided a sample code that demonstrates how to implement Otsu’s thresholding using Python OpenCV. By following the steps outlined in this article, you should be able to perform Otsu’s thresholding on your own images using Python OpenCV, which can be useful in various image analysis and computer vision tasks.