These “American Pixels” are an experiment. Image formats like jpeg (or gif) use compression algorithms to save space, while trying to retain a large fraction of the original information. A computer that creates a jpeg does not know anything about the contents of the image: It does what it is told, in a uniform manner across the image.
My idea was to create a variant that followed in the footsteps of what jpegs do, but to have the final result depend on the original image: in a very direct way the computer algorithm becomes part of the image creation. The idea was to build a hierarchical compression algorithm, where the compression – in effect the pixel size – depends on the information in each uncompressed pixel and its neighbours. So adaptive compression (acomp) is a new image algorithm where the focus is not on making its compression efficient but, rather, on making its result interesting.
[...] What is more, it produces images that have spatial depth: as you zoom in you can see more and more details. acomps are designed for a wall: The viewer has to be able to walk back and forth in front of them.
Basically, the algorithm leaves detail where there needs to be detail, and compresses areas of less detail. By doing this, the resulting image doesn’t look entirely realistic, yet doesn’t look entirely artificial either.