I haven't written about this yet but I don't often see it mentioned: dithering has applications outside of image processing. Any time one needs to create a sequence sampled from a distribution, but would like to do so "evenly" without creating lumps, Floyd–Steinberg is a decent candidate.