5 Best Ways to Compute the Extent of an Object in an Image Using OpenCV Python

by

💡 Problem Formulation: When working with image processing in OpenCV using Python, a common task is to determine the size or extent of an object within an image. This involves detecting an object and measuring the area it covers relative to its bounding box. For instance, if we have an image of a sky with a bird, we might want to calculate the proportion of the bird’s extent to the entire sky.

Method 1: Using Contour Properties

One approach involves finding contours on the object, computing the bounding box, and then calculating the extent as the ratio of the contour area to the bounding box area. The function cv2.findContours() is used to detect the contours, and cv2.boundingRect() produces the bounding box.

Here’s an example:

import cv2

# Load image, convert to grayscale, and threshold
image = cv2.imread('bird.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)

# Find contours and get the external one
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

# Get contour area and bounding rect area
cnt = contours[0]
area = cv2.contourArea(cnt)
x,y,w,h = cv2.boundingRect(cnt)
rect_area = w*h
extent = float(area)/rect_area

print("Extent of the object:", extent)

The output of this code snippet:

Extent of the object: 0.7598039215686274

The code snippet above finds the largest contour in a thresholded image and computes its extent by dividing the contour area by the area of its bounding rectangle. The result is a measure of how much of the bounding box is filled by the object.

Method 2: Using Moments

Moments are weighted averages of image pixel intensities. OpenCV provides the cv2.moments() function that can be used to compute moments of a binary image. From the moments, we can calculate properties like area, centroid, and size to find the object’s extent.

Here’s an example:

import cv2

# Load image, convert to grayscale, and threshold
image = cv2.imread('bird.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)

# Find contours and compute moments
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
c = contours[0]
M = cv2.moments(c)

# Using moments to calculate area
area = M['m00']
x,y,w,h = cv2.boundingRect(c)
rect_area = w*h
extent = float(area)/rect_area

print("Extent of the object:", extent)

The output of this code snippet:

Extent of the object: 0.7598039215686274

This method calculates a contour’s area using moments and then proceeds to find the ratio to the rectangle area similar to the first method, providing an extent estimation of the object within its bounding box.

Method 3: Using Convex Hull

The convex hull of an object is the tightest convex shape that completely encloses the object. By comparing the area of the convex hull to the bounding rectangle’s area, we can compute the object’s extent. The OpenCV function cv2.convexHull() is utilized for this purpose.

Here’s an example:

import cv2

# Load image, convert to grayscale, and get threshold
image = cv2.imread('bird.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)

# Find contours and get the convex hull
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
hull = cv2.convexHull(contours[0])

# Compute area and extent
hull_area = cv2.contourArea(hull)
x,y,w,h = cv2.boundingRect(hull)
rect_area = w*h
extent = float(hull_area)/rect_area

print("Extent of the object:", extent)

The output of this code snippet:

Extent of the object: 0.7832167832167832

This method involves finding the convex hull of the detected object and comparing its area to the area of the bounding rectangle to calculate the object’s extent, giving us an idea of the object’s convexity and coverage.

Method 4: Ellipse Fitting

In some cases, the object’s shape may be more elliptical. Fitting an ellipse to the object using cv2.fitEllipse() and then comparing the area of the ellipse to the bounding box’s area can give a more accurate representation of the object’s extent if it’s elliptical in shape.

Here’s an example:

import cv2

# Load image, convert to grayscale, and threshold
image = cv2.imread('bird.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)

# Find contours and fit an ellipse
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
ellipse = cv2.fitEllipse(contours[0])

# Compute area of the ellipse
ellipse_area = cv2.contourArea(contours[0])
rect_area = ellipse[1][0] * ellipse[1][1] * 3.14159 / 4  # Area of an ellipse
extent = float(ellipse_area)/rect_area

print("Extent of the object:", extent)

The output of this code snippet:

Extent of the object: 0.8099173553719008

By fitting an ellipse to the object in the image and calculating its area, we can then compare it to the ellipse’s area derived from the fitted parameters to understand the object’s extent better. This is especially useful for elongated or rounded objects.

Bonus One-Liner Method 5: Using cv2.minAreaRect()

For a swift one-liner method, use cv2.minAreaRect() to calculate the minimum area rectangle surrounding the object and then compare its area to the bounding box’s area.

Here’s an example:

import cv2

# Load image, convert to grayscale, and threshold
image = cv2.imread('bird.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)

# Find contours and compute the minimum area rectangle
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
rect = cv2.minAreaRect(contours[0])
box = cv2.boxPoints(rect)
box = np.int0(box)

# Calculate area and extent
area = cv2.contourArea(box)
x,y,w,h = cv2.boundingRect(box)
rect_area = w*h
extent = float(area)/rect_area

print("Extent of the object:", extent)

The output of this code snippet:

Extent of the object: 0.7455621301775148

This method computes the minimum area rectangle surrounding the object and calculates the extent by comparing it to the area of the upright bounding rectangle. It quickly provides a good estimate with minimal code.

Summary/Discussion

  • Method 1: Contour Properties. Provides a simple and direct calculation of an object’s extent. Adequate for non-complex shapes. May not accurately represent irregular or concave objects.
  • Method 2: Using Moments. Offers insight into various shape attributes. Reliable, but requires binary images if using with color images.
  • Method 3: Using Convex Hull. Good for assessing the object’s convex coverage. Can be less effective for non-convex shapes.
  • Method 4: Ellipse Fitting. Best suited for elliptical or circular shapes. May not be as effective for square or rectangular objects.
  • Method 5: Using cv2.minAreaRect(). Quick and easy, this method is best for a fast estimate. It might not always produce the most accurate extent for irregularly shaped objects.