Experimental lossless image compression, based on rANS entropy coding (ryg-rans) and traditional prediction and colour space transforms.
Proof of concept, useless for most edge cases.
Encoding:
choh infile.rgb outfile.hoh width height -sN
Where "N" is a number 0-4 (fast-slow). "infile.rgb" must consist of raw 8bit RGB bytes. You can make such a file with imagemagick:
convert input.png -depth 8 rgb:infile.rgb
Encoded size is reported, but no data is written. (Work in progress, it currently writes headers and stuff).
Decoding:
dhoh infile.hoh outfile.rgb
make
That's all you should need. Requires GNU make and gcc. Only tested on 64bit x86.