Pixel sorting is an interesting, glitchy effect which selectively orders the pixels in the rows/columns of an image. It was popularized (possibly invented) by artist Kim Asendorf (processing source code here). The processing script was cryptic and not very hackable, and I felt like something more lightweight was needed, so I wrote my own version in python – more info on GitHub.The earliest reference to pixel sorting I could find on the internet is this paper (PDF) by scientists from Iraq’s Scientific Research Council. They propose using it as a technique for unsupervised classification, to be applied to images from remote sensing satellites. Neat!
Here’s what it looks like in action:
How does it work?
There are two steps involved:
- The rows/columns of the image are split into ‘intervals’ (more on this later).
- The pixels in each interval are rearranged so as to sort them with respect to some property, say, lightness.
Kim Asendorf’s code simply applies this technique twice: first vertically, then horizontally.
The intervals are defined by regions of the image that are too light or too dark – the edges of these regions define the edges of the intervals.
For example, consider:
- The top left is the original image.
- The top right is the image with pixels outside the lightness threshold replaced with black, and the rest filled in with white. Notice that the foam – too light – and the shadows – too dark – are outside the threshold. (Yes, I just discovered en-dashes. Deal with it.)
- On the bottom left is the image with the intervals filled in with random colors – each color represents a different interval.
- On the bottom left is the image with sorted intervals.
How can I pixelsort?
All the images in this article were generated using a pixel sorting script I wrote in python – source here. It shouldn’t be too hard to use, details are in REAMDE.md. If you do anything with it I’d love to see it!
Edit: The script was used to make this music video. Check it out!
