pixelatelib 0.1.0

Learn image processing by building it. A NumPy-based library that exposes both naive and optimized algorithms for side-by-side understanding.

Pixelate

Pixelate is an educational image processing library built to compare naive loop-based algorithms with their vectorized NumPy counterparts. Each operation has a slow and fast implementation so you can explore the performance and readability differences between the two styles.

Features

• Pixel Operations: invert colors, convert to grayscale, and adjust brightness.
• Geometric Transformations: horizontal flip and nearest-neighbor resize.
• Convolution Filters for applying arbitrary kernels.
• Utilities for loading and saving images via Pillow.
• Ready-to-run benchmarks and tests demonstrating correctness and speed differences.

The modules are organized so that every function has a matching fast (vectorized) and slow (naive) version:

pixelate/
  fast/
    filters.py
    pixel_ops.py
    transformations.py
  slow/
    filters.py
    pixel_ops.py
    transformations.py
  utils/
    helpers.py
    image_io.py

Installation

Clone the repository and install the requirements:

pip install -r requirements.txt

You can also install the package directly:

pip install pixelate

The project requires Python 3.10 or newer.

Usage Example

import numpy as np
from pixelate.fast.pixel_ops import invert_color_fast
from pixelate.slow.pixel_ops import invert_color_slow
from pixelate.utils.image_io import load_image, save_image

img = load_image("path/to/image.png")

# Compare slow and fast implementations
slow_result = invert_color_slow(img)
fast_result = invert_color_fast(img)

save_image(fast_result, "out_fast.png")

Check the benchmarks/ folder for scripts that measure execution time for the slow and fast versions of each operation.

Running Tests

Use pytest to run the unit tests:

pytest

License

Pixelate is distributed under the MIT License. See the LICENSE file for details.