I got some time today waiting for the furnace guys to show up (they never did) and managed to figure out a few of the things that were wrong with the image normalization code and with the sync detection code. Here’s the latest automatically generated image from one of the recordings I made earlier.

There are a couple of things about these images which aren’t precisely correct, but overall the rectification works pretty well: the sync detection is working and reasonably reliable in the regions where the image is essentially nose free. With some additional work, I could be even better, and the occasional one pixel glitches in the position of the scanline would likely go away. The normalization isn’t perfect either. But still… pretty neat.