[AI from Scratch] Episode 277: Image Histograms

Recap and Today’s Theme

Hello! In the previous episode, we discussed edge detection techniques, including the Sobel and Canny methods.

Today, we will focus on image histograms, an essential tool in image processing that visually represents the distribution of brightness or color values in an image. In this episode, we will explain the basic concepts of histograms and how to create and utilize them in image analysis.

What is an Image Histogram?

1. Definition of a Histogram

A histogram is a graphical representation of data distribution in the form of a bar chart. In image processing, a histogram shows the distribution of pixel brightness (intensity) or color values. Specifically, the histogram plots the frequency of each brightness level (0 to 255) along the x-axis, while the y-axis shows the number of pixels with that brightness level.

2. Purpose of Histograms

Histograms help analyze the characteristics of an image, such as its contrast, brightness, and color distribution. By interpreting a histogram, you can perform various image processing tasks, such as:

• Contrast Adjustment: Enhancing the image’s brightness or darkness to make it clearer.
• Thresholding (Binarization): Extracting a specific brightness range from an image.
• Image Correction: Adjusting overexposed or underexposed images to make their brightness more uniform.

Grayscale Image Histogram

1. What is a Grayscale Image?

A grayscale image is an image where each pixel represents a brightness level between 0 and 255. In this range, 0 represents black, 255 represents white, and values in between represent various shades of gray.

2. Creating a Grayscale Histogram

To create a histogram for a grayscale image, the frequency of each brightness level is counted. The x-axis represents brightness levels (0–255), and the y-axis represents the number of pixels with each brightness level.

Displaying a Histogram with OpenCV

Here’s an example of how to create and display a grayscale histogram using OpenCV and Matplotlib:

import cv2
import matplotlib.pyplot as plt

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

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

# Display the histogram
plt.plot(hist)
plt.title('Grayscale Histogram')
plt.xlabel('Pixel Value')
plt.ylabel('Frequency')
plt.show()

This code uses cv2.calcHist() to compute the histogram and Matplotlib to visualize it.

3. Interpreting a Grayscale Histogram

By analyzing the histogram, you can understand the characteristics of an image:

• Left-Skewed (Closer to 0): The image contains more dark pixels.
• Right-Skewed (Closer to 255): The image contains more bright pixels.
• Even Distribution: The image has high contrast, with brightness values spread across the full range.

Color Image Histogram

1. Creating a Color Histogram

For color images, a separate histogram is created for each channel—Red, Green, and Blue (RGB). This allows you to visualize the distribution of each color component.

Displaying a Color Histogram with OpenCV

Here’s how to create and display a color histogram for each RGB channel:

# Load a color image
image = cv2.imread('example.jpg')

# Define color channels
colors = ('b', 'g', 'r')
for i, color in enumerate(colors):
    # Compute the histogram for each channel
    hist = cv2.calcHist([image], [i], None, [256], [0, 256])
    # Plot the histogram for the channel
    plt.plot(hist, color=color)

plt.title('Color Histogram')
plt.xlabel('Pixel Value')
plt.ylabel('Frequency')
plt.show()

This code computes and plots histograms for the blue, green, and red channels separately.

2. Interpreting a Color Histogram

A color histogram helps analyze the balance and distribution of colors in an image:

• Strong Presence of a Specific Color: A sharp peak in one color channel.
• Even Distribution of Colors: When all three histograms are balanced, the image contains a mix of colors.
• Color Bias: A histogram skewed towards a particular color indicates a bias towards that color in the image.

Histogram Equalization

1. What is Histogram Equalization?

Histogram equalization is a technique used to improve an image’s contrast by spreading out the pixel values more evenly across the brightness spectrum. It balances dark and light areas, making the image more visually balanced.

2. Implementing Histogram Equalization with OpenCV

You can perform histogram equalization in OpenCV using the cv2.equalizeHist() function. Here’s an example:

# Perform histogram equalization
equalized_image = cv2.equalizeHist(image)

# Display the original and equalized images
cv2.imshow('Original Image', image)
cv2.imshow('Equalized Image', equalized_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

This code equalizes a grayscale image and displays the results before and after equalization.

3. Effects of Histogram Equalization

• Improved Contrast: Enhances details in images with low contrast.
• More Even Distribution: Pixel values are spread more evenly across the brightness range.

However, histogram equalization may not always be effective for all images, and it’s important to evaluate its impact based on the specific image.

Applications of Histograms in Image Processing

1. Contrast Adjustment

By analyzing the histogram, you can adjust the contrast of an image to make it clearer. Enhancing contrast reveals more details and makes features more distinguishable.

2. Thresholding (Binarization)

Histograms help set thresholds for binarization, a process that converts an image into black and white. By selecting a brightness range based on the histogram, you can isolate and highlight specific areas of the image.

3. Color Correction

Color histograms enable color adjustment by balancing colors that are either too dominant or too weak. This helps achieve a more visually appealing and balanced image.

Summary

In this episode, we explored image histograms, an essential tool in image processing. Histograms allow you to visualize the brightness and color distribution in images, making it easier to adjust contrast and perform corrections. By using techniques like histogram equalization, you can enhance the visual quality of images for analysis.

Next Episode Preview

Next time, we will discuss thresholding and the process of converting images into binary form. We will explore various methods for binarization and their applications.

---

Notes

1. Histogram: A graphical representation of data distribution, used in image processing to show pixel brightness or color distribution.
2. Histogram Equalization: A technique to enhance image contrast by spreading out pixel values more evenly.
3. Thresholding: A process that converts an image into binary form by setting a threshold based on pixel brightness.