Automatic detection of scan timing offsets

[benj]

It would be helpful to have some kind of automatic method for detecting correct scan timing offsets for scan lines, to save repeated trial-and-error processing to find the correct offset.

Possible approaches so far include edge detection (Mike), Fourier transforms (Steve) or neural networks (Ben)

[mggr]

Attempted with edge detection (nip2) and measuring total length of edges in image for various offsets, on the assumption that incorrect timing = more wobbles = longer edges. Didn't work :/