Image thresholding (binary, adaptive, Otsu) in Computer Vision

Introduction

Image thresholding helps turn a color or gray image into a simple black and white image. This makes it easier to find shapes or objects.

To separate objects from the background in a photo.

To prepare images for reading text or numbers automatically.

To detect edges or shapes in simple images.

When you want to reduce image details to just black and white.

To clean up noisy images by choosing the right threshold method.

Syntax

Computer Vision

import cv2

# Binary thresholding
_, binary_img = cv2.threshold(src, thresh, maxval, cv2.THRESH_BINARY)

# Adaptive thresholding
adaptive_img = cv2.adaptiveThreshold(src, maxval, adaptiveMethod, thresholdType, blockSize, C)

# Otsu's thresholding
_, otsu_img = cv2.threshold(src, 0, maxval, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

src is the input grayscale image.

For adaptive thresholding, blockSize is the size of the area to calculate the threshold, and C is a constant subtracted from the mean or weighted mean.

Examples

This sets pixels above 127 to white (255) and below to black (0).

Computer Vision

_, binary_img = cv2.threshold(gray_img, 127, 255, cv2.THRESH_BINARY)

Threshold is calculated for each 11x11 block using the mean minus 2.

Computer Vision

adaptive_img = cv2.adaptiveThreshold(gray_img, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 11, 2)

Otsu automatically finds the best threshold value to separate foreground and background.

Computer Vision

_, otsu_img = cv2.threshold(gray_img, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

Sample Model

This code creates a simple image with a bright square on a dark background. It applies three types of thresholding and counts how many pixels are white after each method.

Computer Vision

import cv2
import numpy as np

# Create a simple grayscale image with a dark and bright area
img = np.zeros((100, 100), dtype=np.uint8)
img[25:75, 25:75] = 150  # Bright square in the middle

# Binary thresholding
_, binary_img = cv2.threshold(img, 100, 255, cv2.THRESH_BINARY)

# Adaptive thresholding
adaptive_img = cv2.adaptiveThreshold(img, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 11, 2)

# Otsu's thresholding
_, otsu_img = cv2.threshold(img, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)

# Print threshold values and sum of white pixels to compare
print("Binary threshold sum:", np.sum(binary_img == 255))
print("Adaptive threshold sum:", np.sum(adaptive_img == 255))
print("Otsu threshold sum:", np.sum(otsu_img == 255))

Important Notes

Binary thresholding uses one fixed value for the whole image.

Adaptive thresholding changes the threshold for different parts of the image, useful for uneven lighting.

Otsu's method finds the best threshold automatically by looking at the image histogram.

Summary

Image thresholding turns images into black and white for easier analysis.

Binary thresholding uses a fixed cutoff value.

Adaptive and Otsu methods help when lighting or colors vary across the image.