How to compare histograms of two images using OpenCV Python?

The histograms of two images can be compared using cv2.compareHist() function. This function accepts three input arguments: hist1, hist2, and compare_method. The hist1 and hist2 are histograms of the two input images, while compare_method is a metric to compute the matching between the histograms. It returns a numerical parameter that expresses how well two histograms match with each other.

Comparison Methods

OpenCV provides four different histogram comparison methods:

Steps

To compare the histograms of two images, follow these steps:

• Import the required libraries: OpenCV and Matplotlib

• Read the input images using cv2.imread() function

• Calculate the histograms using cv2.calcHist()

• Normalize the histograms using cv2.normalize()

• Compare histograms using cv2.compareHist() with your chosen method

• Interpret the comparison metric value

Example 1: Basic Histogram Comparison

This example demonstrates comparing two different images using the correlation method :

import cv2
import numpy as np

# Create sample images for demonstration
img1 = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)
img2 = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)

# Calculate histograms for each color channel
hist_img1 = cv2.calcHist([img1], [0, 1, 2], None, [256, 256, 256], [0, 256, 0, 256, 0, 256])
hist_img2 = cv2.calcHist([img2], [0, 1, 2], None, [256, 256, 256], [0, 256, 0, 256, 0, 256])

# Normalize the histograms
cv2.normalize(hist_img1, hist_img1, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX)
cv2.normalize(hist_img2, hist_img2, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX)

# Compare using correlation method
metric_val = cv2.compareHist(hist_img1, hist_img2, cv2.HISTCMP_CORREL)
print(f"Correlation value: {metric_val:.4f}")

Correlation value: 0.0156

Example 2: All Comparison Methods

This example compares histograms using all four available methods :

import cv2
import numpy as np

# Create two sample images
img1 = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)
img2 = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)

# Calculate and normalize histograms
hist_img1 = cv2.calcHist([img1], [0, 1, 2], None, [256, 256, 256], [0, 256, 0, 256, 0, 256])
hist_img2 = cv2.calcHist([img2], [0, 1, 2], None, [256, 256, 256], [0, 256, 0, 256, 0, 256])

cv2.normalize(hist_img1, hist_img1, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX)
cv2.normalize(hist_img2, hist_img2, alpha=0, beta=1, norm_type=cv2.NORM_MINMAX)

# Compare using all methods
correl = cv2.compareHist(hist_img1, hist_img2, cv2.HISTCMP_CORREL)
chisqr = cv2.compareHist(hist_img1, hist_img2, cv2.HISTCMP_CHISQR)
intersect = cv2.compareHist(hist_img1, hist_img2, cv2.HISTCMP_INTERSECT)
bhattacharyya = cv2.compareHist(hist_img1, hist_img2, cv2.HISTCMP_BHATTACHARYYA)

print(f"Correlation: {correl:.4f}")
print(f"Chi-Square: {chisqr:.4f}")
print(f"Intersection: {intersect:.4f}")
print(f"Bhattacharyya: {bhattacharyya:.4f}")

Correlation: -0.0094
Chi-Square: 1.9980
Intersection: 0.0977
Bhattacharyya: 0.7071

Interpreting Results

Understanding the comparison values:

• Correlation: Values closer to 1 indicate better similarity. Negative values suggest inverse correlation.

• Chi-Square: Lower values indicate better similarity. Zero means perfect match.

• Intersection: Higher values indicate better similarity.

• Bhattacharyya: Values closer to 0 indicate better similarity. Values closer to 1 indicate dissimilarity.

Conclusion

Use cv2.compareHist() to quantify histogram similarity between images. Choose the comparison method based on your specific requirements: correlation for general similarity, chi-square for statistical difference, or Bhattacharyya for probabilistic distance.