This GitHub repository contains Mathematica code for the generation of polygonal shapes representing letters of the alphabet. These polygons are derived from low-resolution rasterized fonts, where simple outlines are extracted from the pixelated letters. The resulting data is organized into three directories: whitebackground, nobackground, svgs.
The directory ./whitebackground contains the files {"A.png", "B.png", ..., "Z.png"}. They are all 256x256 png images.
The directory ./nobackground contains the files {"Anobackground.png", "Bnobackground.png", ..., "Znobackground.png"}. They are all 256x256 png images, with transparent pixels instead of white pixels where there is no letter.
The directory ./svgs contains the files {"A.svg", "B.svg", ..., "Z.svg"}. They are vector graphics whose viewBox's are slightly larger than the letters they contain. The viewBox's are not consistent across files.
The letter shapes were created by rasterizing a font at very low resolutions, and extracting simple outlines from the very pixellated letters. With the exception of {Y,R,I,K,Z,F,H,E,W,X,L}, the letters were rasterized at 9x9 pixels. Y was rasterized at 14x14 pixels. R was rasterized at 11x11 pixels. I and K were rasterized at 10x10 pixels. Z, F and H were rasterized at 8x8 pixels. E, W and X were rasterized at 7x7 pixels. L was rasterized at 6 pixels. An illustration of this process can be seen in ./example.png, which contains the letter A rasterized at 9x9 pixels, and its extracted outline overlaid.
One nasty trick is (2) which circumvents BoundaryMeshRegion being AtomQ. Another nasty trick (1) is the contour nesting code which assumes that contours with even depth are black (not 'inner') so may be treated as separate polygons.