How to Count Overlapping Samples in Images with Python (2026)

Counting overlapping samples in images is a common challenge faced by developers working with image processing. In this tutorial, we will explore how to fix Python code to accurately count different sized samples in an image, even when they overlap. We will use the OpenCV library to tackle this problem effectively.

Key Takeaways

• Learn to identify and separate overlapping samples in images using Python.
• Gain skills in using OpenCV for image processing tasks.
• Understand the process of contour detection and image thresholding.
• Develop a method to measure the size of each detected sample.

Introduction

In image processing, counting and measuring the size of different samples within an image is crucial for various applications, such as quality control in manufacturing or biological research. The challenge arises when samples overlap, leading to inaccurate counts and measurements. This tutorial provides a step-by-step guide to fix Python code for counting samples accurately, using OpenCV, a powerful library for computer vision tasks.

Our goal is to develop a Python script that can accurately count and measure the length of overlapping samples in an image, with a focus on a sample image named rods_input.png.

Prerequisites

• Basic knowledge of Python programming.
• Python 3.8 or later installed (tested on Python 3.9.7).
• OpenCV library (version 4.5.3) installed in your Python environment.
• A sample image named rods_input.png for testing.

Step 1: Install Dependencies

First, ensure you have the necessary libraries installed. You can install OpenCV using pip:

pip install opencv-python

Step 2: Load and Preprocess the Image

Load the image and convert it to a grayscale format, which simplifies the process of detecting edges and contours.

import cv2

# Load the image
image = cv2.imread('rods_input.png')

# Convert to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

Step 3: Apply Thresholding

Thresholding helps in separating the objects from the background by converting the image to a binary format.

# Apply GaussianBlur to reduce noise and improve contour detection
gray_blurred = cv2.GaussianBlur(gray, (5, 5), 0)

# Apply thresholding
_, thresh = cv2.threshold(gray_blurred, 127, 255, cv2.THRESH_BINARY_INV)

Step 4: Detect Contours

Use contour detection to find the outlines of the samples. This step is crucial for counting and measuring each sample.

# Find contours
contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

Step 5: Separate Overlapping Samples

To separate overlapping samples, we apply a distance transform and use connected components analysis.

# Apply distance transform
dist_transform = cv2.distanceTransform(thresh, cv2.DIST_L2, 5)

# Normalize the distance transform image
dist_transform = cv2.normalize(dist_transform, None, 0, 1.0, cv2.NORM_MINMAX)

# Threshold to obtain binary image from distance transform
_, dist_thresh = cv2.threshold(dist_transform, 0.5, 1.0, cv2.THRESH_BINARY)

# Convert back to uint8
markers = cv2.convertScaleAbs(dist_thresh * 255)

# Use connected components to label unique objects
num_labels, labels = cv2.connectedComponents(markers)

Step 6: Measure the Length of Each Sample

For each detected sample, calculate its length using the bounding rectangle method.

for i in range(1, num_labels):
    # Create a mask for each component
    component_mask = (labels == i).astype('uint8') * 255

    # Find contours for each component
    component_contours, _ = cv2.findContours(component_mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

    for cnt in component_contours:
        # Calculate the bounding rectangle
        x, y, w, h = cv2.boundingRect(cnt)
        # Calculate the length
        length = max(w, h)
        print(f'Sample {i}: Length = {length}')
        # Draw the bounding rectangle
        cv2.rectangle(image, (x, y), (x+w, y+h), (0, 255, 0), 2)

Common Errors/Troubleshooting

• Blurred Image: Ensure the image is preprocessed with GaussianBlur to improve edge detection.
• Incorrect Contour Detection: Adjust threshold values to better isolate the samples from the background.
• Overlap Not Resolved: Fine-tune the distance transform threshold to better separate overlapping areas.

Conclusion

By following this tutorial, you have learned how to accurately count and measure overlapping samples in an image using Python and OpenCV. This approach is versatile and can be adapted to various types of images and applications, making it a valuable skill for anyone working in image processing.

Frequently Asked Questions

Why is GaussianBlur used before thresholding?

GaussianBlur reduces image noise and details, which helps in better contour detection by smoothing the image.

What is distance transform and why is it used?

Distance transform calculates the distance to the nearest zero pixel, helping in separating overlapping components for accurate counting.

How can I adjust the accuracy of the sample detection?

Adjust the threshold values and the parameters of the distance transform to better match the characteristics of your specific image.