How do I convert a PIL Image into a NumPy array?

python

numpy

pil

image-processing

byAlex Kataev·Oct 9, 2024

Straight to the core, guys. Use np.array() to convert a PIL Image into a NumPy array:

import numpy as np
from PIL import Image

image = Image.open('image.jpg')  # Load your image (the stuff of SSJ Goku!)
image_array = np.array(image)    # BAM! PIL Image to NumPy array (Unleashed your Genki Dama here)

The shape of the output array is (height, width, channels). Keep in mind, channels is 3 for RGB and 4 for RGBA images.

To revert the process, cast the NumPy array back to a PIL Image:

modified_image = Image.fromarray(image_array)
modified_image.save('modified_image.jpg')  # Save the aftermath (post Genki Dama impact!)

Guys! It's crucial to confirm image_array has the correct shape and data type to match the PIL Image mode (e.g. 'RGB' or grayscale 'L').

Journey back to PIL world

Want to rewrite the pixel of an image with a modified NumPy array?

# Assuming 'pix' is your modified NumPy array
image.paste(Image.fromarray(pix))

Just like Goku needs to be in Super Saiyan mode to do his thing, image.paste() requires the input array to match the image size. Similarly, Image.fromarray() needs appropriate color mode and data type alignment.

Tackling color modes and multiverse channels

Transforming just like Goku’s Super Saiyan transformations with different color modes or manipulating channels? Worry not:

# Convert to grayscale before transformation for the ultimate SSJ form!
grayscale_image = image.convert('L')
grayscale_array = np.array(grayscale_image)

# Handle RGBA or convert RGB to BGR (Power up to Super Saiyan Blue!)
bgr_image_array = image_array[...,::-1]  # Flipping channels, just like Goku flips his hair!

In case of channel shuffling, np.rollaxis() or np.transpose() is your Senzu Bean to reshape the array!

Power up with custom functions!

Frequent transformations? Fret not. Create reusable energy blasts, ahem, helper functions!

def pil2array(img):
    """Konverting Kaioken (PIL Image to NumPy array attack!)"""
    return np.array(img)

def array2pil(arr, mode='RGB'):
    """Reversing the Kaioken (NumPy array to PIL Image move!)"""
    return Image.fromarray(np.uint8(arr), mode)

Training with the newest masters (PIL, NumPy)

Always keep an eye on your training partners (Python, PIL, and NumPy versions). The fighting stance ((column-major or row-major format) might change:

image_array = np.array(image)       # Column-major (default in NumPy)
image_array = np.ascontiguousarray(image_array)  # Switch to row-major for a surprise attack!

Join the tournament (Advanced Image processing)

On the battlefield of advanced image processing:

NumPy slicing: An effective move to crop or modify regions (square or otherwise, up to you!)
Develop filtering & transformations techniques: Dokkan Awaken your images!
Feed machine learning models with NumPy arrays: Raise the power levels of your neural networks!

For more secret techniques, opt for more research on advanced PIL and NumPy methods.

Ensuring a fair battle (Potential pitfalls)

Rub your Dragon Balls and watch out for:

Memory usage: Large arrays can exhaust RAM - work in small batches to avoid a crash.
Data types: Be careful with 'I' and 'F' modes - they create arrays of int32 and float types respectively.
Coordinate systems: PIL and NumPy have different conventions for pixel coordinates.

explain-codes / Python / How do I convert a PIL Image into a NumPy array?

Linked

How to Crop an Image in OpenCV Using Python



How do I add an extra column to a NumPy array?



Convert NumPy array to Python list



How do I resize an image using PIL and maintain its aspect ratio?



Programmatically generate video or animated GIF in Python?

