OpenCV

OpenCV (Open Source Computer Vision Library) is the standard library for real-time computer vision, image processing, camera calibration, and deep learning model inference.

When to Use

• Real-Time Computer Vision & Video Processing: Frame-by-frame analysis of camera feeds, robotics, and industrial automation.
• Image Preprocessing for Deep Learning: Resizing, normalizing, thresholding, and color-space conversions.
• Edge, Contour & Shape Detection: Canny edge detection, Hough transforms, and bounding box polygon calculations.
• Feature Matching & Homography: Aligning perspectives and matching keypoints with ORB and SIFT.

Quick Start

import cv2

# Load image, convert to grayscale, apply Gaussian blur and Canny edge detection
img = cv2.imread("input.jpg")
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
blurred = cv2.GaussianBlur(gray, (5, 5), 0)
edges = cv2.Canny(blurred, threshold1=50, threshold2=150)

cv2.imwrite("edges.jpg", edges)

Core Concepts

Image Preprocessing, Thresholding & Contours

Detecting object shapes and drawing bounding boxes:

import cv2
import numpy as np

# Load image in grayscale
image = cv2.imread('circuit_board.jpg')
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

# Apply Gaussian blur to reduce high-frequency noise
blurred = cv2.GaussianBlur(gray, (5, 5), 0)

# Canny edge detection
edges = cv2.Canny(blurred, threshold1=50, threshold2=150)

# Find contours in the binary edge image
contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

output = image.copy()
for cnt in contours:
    area = cv2.contourArea(cnt)
    if area > 100: # Filter small noise specks
        x, y, w, h = cv2.boundingRect(cnt)
        cv2.rectangle(output, (x, y), (x + w, y + h), (0, 255, 0), 2)
        cv2.putText(output, f"Area: {int(area)}", (x, y - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 1)

cv2.imwrite('detected_components.jpg', output)

Color Space Filtering with HSV Masking

Isolating colored objects independent of scene brightness:

# Convert to Hue-Saturation-Value (HSV)
hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)

# Define range for blue color in HSV
lower_blue = np.array([100, 150, 50])
upper_blue = np.array([140, 255, 255])

# Threshold image to obtain binary mask
mask = cv2.inRange(hsv, lower_blue, upper_blue)

# Bitwise-AND mask and original image
blue_segmented = cv2.bitwise_and(image, image, mask=mask)
cv2.imwrite('blue_segmented.jpg', blue_segmented)

Video Stream Processing Loop

Reading frames continuously from a camera feed or video file:

cap = cv2.VideoCapture(0) # 0 = default webcam

while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    # Process frame
    gray_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    cv2.imshow('Live Monitor', gray_frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

Common Patterns

Real-Time Video Stream Processing with FPS Throttling

Problem: Processing high-resolution webcam frames in Python causes frame buffering lag.

Solution: Process frames in a tight loop with resolution resizing:

cap = cv2.VideoCapture(0)
# Set capture resolution
cap.set(cv2.CAP_PROP_FRAME_WIDTH, 640)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 480)

while cap.isOpened():
    ret, frame = cap.read()
    if not ret: break

    # Process frame
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    cv2.imshow('Live Stream', gray)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

Best Practices

Do:

• Remember OpenCV stores images in BGR format by default; convert with cv2.COLOR_BGR2RGB before Matplotlib/PyTorch.
• Always release hardware resources (cap.release()) and destroy windows (cv2.destroyAllWindows()).
• Use HSV or LAB color spaces rather than RGB when performing color-based segmentation under varying lighting.
• Apply Gaussian or bilateral blurring prior to edge detection to suppress noise.

Don't:

• Perform heavy deep learning inferences sequentially inside the video capture loop; use background threads.
• Hardcode kernel sizes with even numbers; blurring and morphological kernels should be odd (e.g. 3x3, 5x5).
• Assume cv2.imread() succeeded without checking if the returned array is not None.

Troubleshooting

References

• OpenCV Documentation