The methodology seems obvious - interpret every three bytes in the text as a pixel in an image, compress the image, then reinterpret each (post-compression) pixel as three ASCII characters.
Thanks. Great explanation. Considering that the ASCII character map is arbitrary and there are so many great ways to test lossy compression algorithms( PSNR to A/B testing), this seems like a pointless exercise, but at least I see what the author was trying to do now.
If I wanted to demonstrate lossy compression, I'd use Image Subtraction plugin in GIMP, not ASCII.
This is not about OCR.