A program that applies a .gpl palette to an image by finding the nearest color in the palette for each pixel using the CIECAM02-UCS color space by default (others are available as well). This provides more accurate-to-eye results than can be achieved with a run-of-the-mill graphics editor.
Requires Python 3.6+ and the following packages to run from source:
- numpy
- colorspacious
- imageio
- psutil
- requests
- multiprocess
Or, on Windows 64-bit, there are no additional requirements when using the executable version (found in releases).
If you want to run from source, just download the source from releases and unzip it (recommended), or clone with git clone https://github.com/PureAsbestos/Image-palettizer.git. Then install dependencies like this: pip3 install -r requirements.txt. Run main.py to get started.
To run from executable on Windows 64-bit, simply download the single-executable version from releases, or, for slightly faster startup times, use the portable installation.
To begin, you’ll need two things: a palette in .gpl format, and an image (png, jpeg, etc.) When you run main.py (or the executable), you’ll have some options on screen, allowing you to select specifically what you want to do (dithering, etc.). You can click the Browse buttons to provide paths for the palette and image (it's worth noting that the image location can be a URL). You can also click the Get palette from Lospec button to quickly grab a palette from Lospec's excellent palette list. When you are ready to begin, click Apply. You should see some loading bars, and then a preview window will pop up, allowing you to save your image (Note that the preview image may be of reduced quality, but this will not affect the final output). Click Save As... to choose where to save the file (and what to call it), then click Save to save it. See Features for specific capabilities.
Palette can be found here: Dawnbringer's 16 color palette
On the left: original image
On the right: image palettized with Aseprite (Endesga 32 palette, Bayer 2x2)
On the left: original image
On the right: image palettized with Image Palettizer (Endesga 32 palette, Bayer 2x2)
Besides the core features that have been discussed, this program offers:
- The ability to dither with 9 common error-diffusion methods
- The ability to dither with 4 sizes of Bayer matrix (2x2, 3x3, 4x4, 8x8)
- Integration with Lospec for easy retrieval and use of color palettes
- The ability to work with large images (note that extremely wide images may take up too much RAM)
- Partial parallelism for faster run time on multi-core systems
- Improve code commentary
- Add the ability to run with command-line arguments
- Add a Bayer matrix generator
- Add the ability to work with multiple images, videos, and animated gifs (encode videos as gifs?)
- Add an optimized color palette generator
Add GUI examples- The ability to dither using a loaded image as the threshold matrix (this is especially useful for dithering with a blue-noise texture, but any texture can be used)