Pillow in Python: How to Process Images? (2024)

In Python programming, the importance of handling images efficiently cannot be overstated. Whether you’re creating graphics for data visualization, or delving into deep learning projects, the ability to manipulate and process images is often crucial. This is where Pillow, a powerful Python Imaging Library (PIL) fork, comes into play.

Pillow serves as a comprehensive library for opening, manipulating, and saving many different image file formats. It builds upon the PIL library, offering enhancements, bug fixes, and support for Python 3.x. Pillow provides various functionalities, ranging from basic image manipulation tasks like resizing and cropping to more advanced operations such as filtering, blending, and image enhancement.

The applications of Pillow are diverse and extensive. Here are some common use cases:

• Image Loading: Pillow is mainly used to load, decode and deliver images for other frameworks e.g. Keras, PyTorch or TensorFlow e.g.

im = Image.open("/path/to/image.jpg")

• Image Processing: The library allows users to perform various image processing tasks such as resizing, rotating, and cropping images to suit specific requirements e.g.

im_rotated = im.rotate(90)

• Image Filtering: It provides functionalities to apply different filters to images, including blurring, sharpening, and edge detection, enhancing their visual quality e.g.

enhancer = ImageEnhance.Sharpness(im) 
factor = 2 # increase sharpness 
im = enhancer.enhance(factor)

• Image Enhancement: Pillow facilitates adjusting image properties like brightness, contrast, and color balance, enhancing the overall appearance of images e.g.

enhancer = ImageEnhance.Contrast(im) 
factor = 0.5 # decrease contrast 
im = enhancer.enhance(factor)

• Image Conversion: Pillow supports converting images from one file format to another, making compatibility across different platforms and applications convenient e.g.

im.save("img.bmp")

Pillow users make bugs during coding, as with any other library. Fortunately, Pillow has an active community and thorough documentation, which are invaluable resources for debugging.

• Documentation: The Pillow documentation provides comprehensive guides, tutorials, and references, aiding users in understanding the library’s functionalities and troubleshooting common issues. However, debugging based only on theoretical knowledge is difficult and time-consuming.
• Community Support: Users can seek help and guidance from the Pillow community through forums, mailing lists, or platforms like Stack Overflow. Experienced users and contributors often assist in resolving issues and sharing best practices. Nevertheless, this solution is based on the experience of others and the cases found on the Internet may not reflect 100% of the problem that one may encounter. Especially when it is a bug that is not standard and simple to solve.
• Debugger: Running the code in debug mode allows users to walk through the code step-by-step and check all variables at any stage. It’s a great approach, but it’s challenging to understand such complex variables as images by just looking at the numbers.

Pillow is an image library, so it would be nice to check results visually. Debugging images effectively involves a combination of techniques and tools to identify and resolve issues. Here are some strategies:

Save the picture to the disk

The first possible method for visual debugging entails saving the image to your local storage. To execute this, insert the subsequent code at the point where the variable you wish to inspect undergoes alteration:

img.save("/path/to/image.png")

After implementation, run your application until you reach the aforementioned line. Then, check the saved image in the IDE or in the directory you specified as the argument.

Include a code snippet for image display

This approach is quite similar to the previous one, albeit with a different output display mechanism. In this method, the image isn’t saved to the hard disk but is instead shown in a popup window on the screen.

To utilize this method for displaying a Pillow image, simply include the following code snippet at the line where you intend to observe the image after relevant changes have been applied:

img.show()

Run the code for displaying the image in the debugger

The alternative, for those who use PyCharm, involves utilizing its native debugger, providing a cleaner debugging process compared to the aforementioned methods. It eliminates the need to add any extra code fragments that won’t be necessary in the production stage.

All code is inputted within the “Evaluate Expression” window, accessed via the keyboard shortcut Alt+F8 (Windows and Linux) or ⌥F8 (macOS). This eliminates the requirement to write any code directly into the program during debugging. Once the evaluator is closed, the entered code disappears.

To display the image using this method, simply input the following code (same as previously mentioned) into the evaluator:

img.show()

Use a visualization tool

The Debug Image Viewer plugin offers versatile functionality, presenting data in two distinct modes: pop-up and dialog. By adjusting this setting, users can explore and determine the mode that best suits their preferences. This plugin provides essential details such as the variable name, image size, depth, and data type, enhancing the user’s understanding of the displayed image. Additionally, users can conveniently save images in PNG format, apply colormap, or display channel by channel. The support for Pillow (TensorFlow and PyTorch also) is available for PyCharm as a premium feature (but you can start a 30-day free trial to test it :)). You can check all the features here.

In conclusion, Pillow stands as an indispensable tool for image processing in Python, offering a rich set of functionalities and ease of use. By understanding its capabilities, leveraging community support, and employing effective debugging techniques, developers can harness the full potential of Pillow to manipulate and enhance images for a myriad of applications including Deep Learning.

Disclaimer: The author of this article is also the author of the Debug Image Viewer plugin.