Deal with wiggle traces

[kwinkunks]

Possibly very hard indeed. Here's a real challenge...

https://files.slack.com/files-pri/T02SXUB1V-F0B6AL89X/pasted_image_at_2015_09_23_09_10_am.png

(private image shows scanned seismic data with wiggle traces.)

[EvanBianco]

FWIW, I’ve seen a whack of TIFFs digitized by other companies, (who knows who), who end up making the equivalent of 30 some odd “traces” for each wiggle trace. It’s the notion that they aren’t actually digitizing the wiggle function values, but rastering the image straight up. I’m inclined to think that doing a sparse decimation of the images at the exact trace location (at the zero crossing of the wiggle) would achieve essentially a two bit image (black or white), but that time series would more like variable density than oversampling. One would have to deal with the vertical lines, perhaps by slightly offsetting. In sum, I don’t think anyone actually “digitizes” seismic data properly (is hard), rastering a wiggle plot…sure. It occurs to me instead of dealing with all the lines and wiggles when digitizing, maybe better to smooth the image (2Dconv) once digitized? And then take every 30th “trace”.