No real instructions, meant for a Zelda project I'm doing with @TheUltimatum It runs decent now. It scans tiles from spritesheets, then matches 'em up, and stamps 'em out as a full map. Had some trouble for a moment when I raised the resolution. Was using 9 samples per tile, raised that to 25. Think it was trying to be too perfect and missing potential matches. So I reprogrammed it to give its best guess from basically 2D point cloud data.
The Tile-scan part of it does OK so far. Letting it run right now to collect-the-dots. Redid the code for matching the point cloud data with the maps. It's fairly accurate now, and should do well enough for my purposes, so I'm happy. I've got another thought on speeding it up a bit, by keeping a list of the last 20 tiles stamped. I'll do that later on tonight perhaps. It keeps each scanned point in a separate list, so it can then compare the tile location with the raw data, and pick the one that gets the most matches. say 24 out of 25 points would be fine... It was multiplying the RGB values into decimal, then joining those 9 points into one distinct number. Essentially checking against a hash table, tho that way was missing too many. Problem is, I don't know how pixel-perfect Scratch's scan thing is. It was a 2x2 pixel, 'cuz that's what I had in the previous app. Who's to say which color it was actually scanning if it's touching more than one... Gonna try 1x1 and see if we get better results. Might have to zoom in more, then move the map sprite around to be under the scanner pixel.
