Cody Claus
gumroad
Adopt Pixels - Aseprite Extension
🕑 Added 2022-08-09 16:09:24 +0000 UTCYou'll get an .aseprite-extension file containing the Adopt Pixels extension and some instructions on how to add it to Aseprite.
Adopt Pixels is a denoise algorithm that removes "orphan pixels" in a palette and style friendly way.
Typical denoise methods blur or mix pixels to create a smoother image, but with pixel art being at such a small scale this tends to completely destroy sprites. My method just looks at the pixels nearby, and if there aren't enough of the same color pixel it changes the color to the most common of the surrounding pixels. This helps remove orphan pixels and noise, while leaving most details intact. Settings can be changed to remove more pixels, or repeat the process multiple times.
Sea Monster denoise example
Here you can see the algorithm going through multiple "Thresholds" and "Iterations" as it progresses. You can also see the method does a decent job of keeping the majority of the shape and structure, even at very high denoise levels.
Mushroom House denoise example
This example shows four levels of denoising, increasing from left to right, top to bottom. Iterations was set to 1 for each image (that's why the lower images still have orphans).
The "Adopt Pixels" option is located under Edit -> FX -> Adopt Pixels.
Adopt Pixels option location in Aseprite
You will be asked for "Orphan Threshold" and "Denoise Iterations".
Threshold is the number of adjacent pixels of the same color required to keep the subject pixel.
Iterations is how many times the script will repeat the algorithm. This option is helpful for larger thresholds to reach a stable denoised image. (For example, the mushroom house takes 6 iterations on a threshold of 3 to reach a fully denoised image.)
If you have any issues with the extension please reach out to me by joining the Retro Diffusion Discord server: https://discord.gg/baCV7nJn9g
Thanks for using the extension! If you have any issues or suggestions contact me on Discord: Astropulse#9690
