Post #4258505

@vwbusguy@mastodon.online When you say "scans," is this from a (flat bed) scanner or are they from a camera photo? If it's the former, it's a lot easier since there's uniform lighting and no lens distortion. A semi-automatic script could handle it by sampling/summing colours in each grid area. It would have to be guided by a manual/interactive part that selects the area of interest, its orientation in the scan, and what the dimensions of the area are in terms of grid block. There are more complicated ways of going about it if the images aren't as uniform, but that's getting into image processing, feature extraction and some sort of neural network/belief network/machine learning, though. If you have higher fidelity/less noisy inputs, though, a lot of it is unnecessary. If you're doing enough of this sort of conversion, it probably is worth automating the part that detects scale, position and rotation. This is mostly Image Processing 101. I'm sure there are people who use this stuff every day and who know a lot more about it than me. They'll probably (hopefully) be able to point you to an implementation (rather than me hand-waving about how it could/should work in theory...).