#imageprocessing — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #imageprocessing, aggregated by home.social.
-
Part 3 of Type-Safe Computer Vision is up: Climbing the Pyramid.
Shrink an image to find objects at any size and you inherit three traps: odd sizes that never round-trip, two things both called "scale", and half a pixel of drift per level.
In fovea 0.5 each pyramid level carries its own pixel distance and origin offset, so a detection lifts back into base-image coordinates with one call, level.to_base(peak).
https://karhunen-loeve.github.io/03-pyramids-and-scale-space.html