[priomsrb]

I wonder if the end result could be fed into a compression algorithm. The algorithm would need to have a low data overhead as the snapshot sizes are around 32 kilobytes (256 kilo bits).

It would be interesting to know what compression algorithms work well for this kind of application. One example I can think of is Run Length Encoding. Or perhaps an algorithm that is trained on example datasets and finds bit patterns that repeat often.