Interesting question.
If your image has 256 levels of gray (8-bit), and you reduce it to black and white, you need to first enlarge it by 16-times in each direction to contain the same amount of data. (16*16 = 256.)
And the process should be reversible: from the large monochrome image, shrinking it back should restore the original.
For example, g.png is an 8-bit grayscale image, size 267x233 pixels. We make g2.png, with 4272x3728 pixels that are each black or white. Then we reverse the process.
Code: Select all
%IM%convert xc:black xc:white +append bw.png
%IM%convert g.png -resize 1600% -remap bw.png g2.png
%IM%convert g2.png -scale "267x233^!" +depth 8 g3.png
%IM%compare -metric RMSE g3.png g.png NULL:
262.524 (0.00400586)
We have restored the original, to about 0.4% RMSE, which is all we can expect for 8-bit images.
EDIT: I should have said, those commands are Windows CMD syntax. Adjust as required for other shells.
EDIT2: I don't know what you mean about potrace. Perhaps I misunderstand what you want.