Computer Vision Libraries (OpenCV & Pillow)

Introduction

Working with images in Python usually means reaching for one of two libraries: OpenCV, a heavyweight computer vision toolkit built for complex processing pipelines and object detection, or Pillow, a lightweight library for everyday tasks like resizing, cropping, and converting image formats.

This lesson covers both, with an edge-detection example in OpenCV and a resize-and-save example in Pillow.

What is OpenCV?

OpenCV (Open Source Computer Vision Library) is a comprehensive toolkit for image and video processing, originally written in C++ with Python bindings. It covers everything from basic filtering and edge detection to face detection, object tracking, and the preprocessing pipelines that feed into deep learning models for computer vision.

The pip package is named opencv-python, but it is imported in Python as cv2 — a naming holdover from OpenCV's original C++ version 2 API.

pip install opencv-python

Example: Edge Detection with OpenCV

Edge detection highlights the boundaries between regions of an image, and is a common early step in object detection and image analysis pipelines. The example below loads an image, converts it to grayscale, and applies the Canny edge detector.

import cv2

# Load an image from disk
image = cv2.imread("photo.jpg")

# Convert to grayscale, since edge detection does not need color information
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

# Detect edges using the Canny algorithm
edges = cv2.Canny(gray, threshold1=100, threshold2=200)

# Save the result to a new file
cv2.imwrite("photo_edges.jpg", edges)

print(f"Original shape: {image.shape}")
print(f"Edges shape: {edges.shape}")

The original image has 3 color channels (blue, green, red — OpenCV loads images in BGR order, not RGB), while the edges output has just one channel, since it is a black-and-white map of where edges were detected.

What is Pillow?

Pillow (imported as PIL, a continuation of the older Python Imaging Library) is a much simpler library focused on everyday image tasks: opening, resizing, cropping, rotating, converting between formats, and basic drawing. It does not include OpenCV's advanced computer vision algorithms, but for straightforward image manipulation it is faster to write and easier to read.

pip install Pillow

Example: Resizing and Converting with Pillow

from PIL import Image

# Open an image
img = Image.open("photo.jpg")
print(f"Original size: {img.size}, format: {img.format}")

# Resize to a thumbnail while preserving aspect ratio
img.thumbnail((300, 300))

# Convert to grayscale and save as PNG
grayscale = img.convert("L")
grayscale.save("photo_thumbnail.png")

print(f"Resized size: {img.size}")

Pillow vs OpenCV

Common Mistakes

Best Practices

Frequently Asked Questions

Key Takeaways

Summary

OpenCV and Pillow cover the two ends of the image-processing spectrum in Python: OpenCV for serious computer vision pipelines, and Pillow for quick, everyday image manipulation. Most projects that touch images end up using at least one of the two, and larger ones often use both.