Image I/O with OpenCV and PIL (Pillow) in Python

Image I/O (Input/Output) is the fundamental process of reading from and writing to image files using programming libraries. In Python, two of the most prominent libraries for this task are OpenCV and PIL (Pillow). A solid understanding of these libraries is essential for anyone involved in image processing, computer vision, or machine learning.

This documentation explores the core methods for reading, displaying, and saving images using both OpenCV and PIL (Pillow), including syntax, practical examples, and common use cases.

Why Use OpenCV and PIL (Pillow)?

OpenCV (Open Source Computer Vision Library)

OpenCV is a powerhouse for real-time computer vision tasks. It's highly efficient and optimized for high-performance operations on images and videos, making it ideal for computationally intensive applications.

PIL (Pillow)

PIL, now actively maintained as Pillow, is a versatile Python Imaging Library. It's renowned for its simplicity, ease of use, and effectiveness in general-purpose image manipulation, such as resizing, format conversion, and basic editing.

1. Image I/O Using OpenCV

Installation

To install OpenCV in Python, use pip:

pip install opencv-python

Reading an Image

import cv2

# Read the image from file
image_path = "example.jpg"
image = cv2.imread(image_path)

# Check if the image was loaded successfully
if image is not None:
    print(f"Image '{image_path}' loaded successfully.")
    print(f"Image shape: {image.shape}") # (height, width, channels)
    print(f"Image data type: {image.dtype}")
else:
    print(f"Failed to load image from '{image_path}'.")

Explanation:

• cv2.imread(path): This function loads an image from the specified file path.
• Return Value: It returns a NumPy array representing the image. By default, OpenCV reads images in BGR (Blue, Green, Red) color channel order, which is a common convention in computer vision.

Displaying an Image

import cv2

# Assuming 'image' is a valid NumPy array loaded by cv2.imread()

if image is not None:
    window_name = "Displayed Image (OpenCV)"
    cv2.imshow(window_name, image)

    # Wait indefinitely until a key is pressed
    # 0 means wait for any key. Pass a positive integer for milliseconds.
    cv2.waitKey(0)

    # Destroy all OpenCV windows
    cv2.destroyAllWindows()
else:
    print("Cannot display image: Image not loaded.")

Explanation:

• cv2.imshow(window_name, image): This function displays the image in a GUI window. The first argument is the name of the window, and the second is the image data (NumPy array).
• cv2.waitKey(0): This is crucial for displaying the image. It waits indefinitely until any key is pressed on the keyboard. If you provide a positive integer (e.g., cv2.waitKey(5000)), it will wait for that many milliseconds before proceeding. This is often used in video processing to display frames sequentially.
• cv2.destroyAllWindows(): This function closes all the OpenCV windows that were opened.

Saving an Image

import cv2

# Assuming 'image' is a valid NumPy array

output_path = "output_image.jpg"
if image is not None:
    success = cv2.imwrite(output_path, image)
    if success:
        print(f"Image saved successfully to '{output_path}'.")
    else:
        print(f"Failed to save image to '{output_path}'.")
else:
    print("Cannot save image: Image not loaded.")

Explanation:

• cv2.imwrite(filename, image): This function writes the image data to a file. The filename argument determines the output format based on its extension (e.g., .jpg, .png).

2. Image I/O Using PIL (Pillow)

Installation

To install Pillow (the modern fork of PIL), use pip:

pip install pillow

Reading an Image

from PIL import Image

# Open the image file
image_path = "example.jpg"
try:
    pil_image = Image.open(image_path)
    print(f"Image '{image_path}' opened successfully with Pillow.")
    print(f"Image format: {pil_image.format}")
    print(f"Image size (width, height): {pil_image.size}")
    print(f"Image mode: {pil_image.mode}") # e.g., 'RGB', 'L' (grayscale), 'RGBA'
except FileNotFoundError:
    print(f"Error: Image file not found at '{image_path}'.")
except Exception as e:
    print(f"An error occurred while opening the image: {e}")

Explanation:

• Image.open(path): This function opens an image file and returns a PIL Image object.
• Image Object Attributes:
  • .format: The format of the image file (e.g., 'JPEG', 'PNG').
  • .size: A tuple representing the width and height of the image in pixels.
  • .mode: The pixel format of the image (e.g., 'RGB' for true color, 'L' for grayscale, 'RGBA' for RGB with alpha channel).

Displaying an Image

from PIL import Image

# Assuming 'pil_image' is a valid PIL Image object

try:
    pil_image.show()
    print("Attempting to display image using the default system viewer.")
except Exception as e:
    print(f"Could not display image: {e}")

Explanation:

• pil_image.show(): This method displays the image using your system's default image viewer. It does not create an in-application window like cv2.imshow().

Saving an Image

from PIL import Image

# Assuming 'pil_image' is a valid PIL Image object

output_path = "output_image.png"
try:
    pil_image.save(output_path)
    print(f"Image saved successfully to '{output_path}'.")
except Exception as e:
    print(f"Failed to save image to '{output_path}': {e}")

Explanation:

• pil_image.save(filename): This method saves the image to a file. The output format is automatically inferred from the file extension provided in filename. Pillow supports a wide range of image formats.

Key Differences Between OpenCV and PIL (Pillow)

Interconversion Between OpenCV and PIL

It's often necessary to convert images between OpenCV's NumPy array format and Pillow's Image object.

From PIL to OpenCV

When converting a PIL image to an OpenCV NumPy array, remember to swap the color channels from RGB to BGR.

import numpy as np
from PIL import Image
import cv2

# Assuming 'pil_image' is a PIL Image object opened with Image.open()

# Convert PIL Image (RGB) to NumPy array
pil_array = np.array(pil_image)

# Convert RGB to BGR for OpenCV compatibility
cv_image = cv2.cvtColor(pil_array, cv2.COLOR_RGB2BGR)

print("Converted PIL image to OpenCV format (BGR NumPy array).")

From OpenCV to PIL

When converting an OpenCV NumPy array back to a PIL Image object, swap the color channels from BGR to RGB.

import cv2
from PIL import Image
import numpy as np

# Assuming 'cv_image' is a NumPy array loaded by cv2.imread()

# Convert BGR to RGB for PIL compatibility
rgb_image_array = cv2.cvtColor(cv_image, cv2.COLOR_BGR2RGB)

# Convert NumPy array to PIL Image object
pil_image_from_cv = Image.fromarray(rgb_image_array)

print("Converted OpenCV image to PIL format (RGB Image object).")

When to Use Which Library?

• Use OpenCV when:

  • You are working with real-time video streams.
  • Your project involves complex computer vision tasks like object detection, facial recognition, or feature matching.
  • You need high-performance image processing operations and optimizations.
  • You are performing mathematical operations on image pixels using NumPy.

• Use PIL/Pillow when:

  • You need to work with a wide variety of image file formats (e.g., converting between JPEG, PNG, GIF, TIFF).
  • Your tasks involve basic image editing: resizing, cropping, rotating, color adjustments, applying simple filters.
  • You need to create images from scratch or draw on existing images.
  • You need to display images simply without complex GUI integration.

Conclusion

Mastering image I/O with OpenCV and PIL (Pillow) is a critical skill for Python developers in imaging domains. OpenCV provides the performance and advanced functionalities required for cutting-edge computer vision, while Pillow offers unparalleled simplicity and ease of use for general image manipulation. You can leverage either library independently or combine their strengths for comprehensive image processing workflows.

SEO Keywords

• Load images with OpenCV Python
• Display images using OpenCV
• Save images in Python OpenCV
• Load and display images with Pillow
• Image processing Python tutorial
• OpenCV vs Pillow image handling
• Python image I/O examples
• Save image file formats Python
• Display images in system viewer Python
• Python libraries for image processing
• PIL vs OpenCV image conversion
• BGR vs RGB color order Python

Interview Questions

• How do you load an image using OpenCV in Python?
• What is the primary difference between image loading in OpenCV and Pillow regarding the returned data type?
• How does OpenCV represent color channels compared to Pillow?
• How can you display an image using OpenCV, and how is it different from displaying an image with Pillow?
• What is the purpose of cv2.waitKey() in OpenCV image display?
• How do you save an image to disk using both OpenCV and Pillow?
• What are some of the image formats that Pillow can save, and how does it determine the format?
• Why might you choose Pillow over OpenCV for specific image manipulation tasks?
• What type of object does Pillow return when loading an image, and how does it differ from OpenCV's return type?
• How would you handle the scenario where an image fails to load in OpenCV?