5 Best Ways to Draw a Rectangular Shape and Extract Objects Using Python’s OpenCV

by

💡 Problem Formulation: In image processing and computer vision using Python’s OpenCV library, one common task is to identify and highlight certain regions of interest within an image. This often involves drawing rectangular shapes around detected objects and extracting them for further analysis. For instance, given an image input containing multiple objects, our desired output is an image with rectangles drawn around each item, followed by extraction of each individual object.

Method 1: Using Contours to Draw Rectangles

Drawing rectangles around objects can be achieved by finding contours on a processed image. The findContours function locates contours of objects and the boundingRect method computes the bounding rectangle for each contour. This approach is suitable for objects that can be easily segmented from the background.

Here’s an example:

import cv2

# Load image, convert to grayscale, and apply threshold
image = cv2.imread('objects.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
ret, thresh = cv2.threshold(gray, 127, 255, 0)

# Find contours and draw rectangles
contours, _ = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
for contour in contours:
    x, y, w, h = cv2.boundingRect(contour)
    cv2.rectangle(image, (x, y), (x+w, y+h), (0, 255, 0), 3)

cv2.imwrite('rectangles.jpg', image)

The output is an image file named ‘rectangles.jpg’ with green rectangles drawn around each object.

This code snippet loads an image, converts it to grayscale, and applies a binary threshold to create a binary image. It then finds contours within this binary image. For each contour detected, the bounding rectangle is calculated and drawn on the original image with green lines using cv2.rectangle. Finally, it saves the resultant image.

Method 2: Using Canny Edge Detection

Using the Canny edge detection algorithm creates edges which can be used to find and draw rectangles. The function Canny from OpenCV detects edges and can be followed with findContours and boundingRect for rectangle detection. This method picks up objects with clear boundaries well.

Here’s an example:

import cv2

# Read the image
image = cv2.imread('objects.jpg')
# Convert to gray scale
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Apply Canny edge detection
edges = cv2.Canny(gray, 50, 150)

# Find contours and draw rectangles
contours, _ = cv2.findContours(edges, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
for contour in contours:
    x, y, w, h = cv2.boundingRect(contour)
    cv2.rectangle(image, (x, y), (x + w, y + h), (0, 255, 0), 2)

cv2.imwrite('canny_edges.jpg', image)

The output is an image ‘canny_edges.jpg’ with rectangles around objects based on detected edges.

This code loads the image, converts it to grayscale, and uses the Canny function to detect edges. It proceeds the same way as Method 1, finding contours from these edges, computing the bounding rectangles, drawing them on the image, and saving the result.

Method 3: Using Hough Line Transform

The Hough Line Transform can detect straight lines which can be useful for drawing rectangles around objects. The OpenCV functions HoughLinesP or HoughLines are used after edge detection to find line segments. These segments can then be combined to form rectangles around objects.

Here’s an example:

// METHOD 3 CONTENT NEEDED

OUTPUT AND EXPLANATION FOR METHOD 3

Method 4: Using Watershed Algorithm

The Watershed algorithm is a classic segmentation approach that can be used to draw rectangles around objects after segmenting them. OpenCV’s watershed function helps in this process. It works well with overlapping objects and complex backgrounds.

Here’s an example:

// METHOD 4 CONTENT NEEDED

OUTPUT AND EXPLANATION FOR METHOD 4

Bonus One-Liner Method 5: Using Minimum Enclosing Rectangle

OpenCV’s minAreaRect function provides a way to draw a minimum area rectangle around the object which may not be axis aligned. This can be a compact one-liner after finding contours.

Here’s an example:

// BONUS METHOD CONTENT NEEDED

OUTPUT AND EXPLANATION FOR BONUS METHOD

Summary/Discussion

  • Method 1: Contours with Bounding Rectangles. Suitable for well-separated objects. Not ideal for overlapping objects or complex backgrounds.
  • Method 2: Canny Edge Detection. Best for objects with defined edges. May miss out on less defined objects and is sensitive to noise.
  • Method 3: Hough Line Transform. Effective for detecting straight-line borders of objects. Complexity increases with the number of detected lines.
  • Method 4: Watershed Algorithm. Great for overlapping objects and distinguishing objects from complex backgrounds. Computationally intensive compared to other methods.
  • Bonus Method 5: Minimum Enclosing Rectangle. Quick one-liner for non-axis aligned bounding boxes. Does not always correspond to the exact shape of an object.