I have a question: why do you need a PNG? How are you going to use it? Why do you need to superpose PNG and JSON?
In VG, when a user requests a map of a specific size, the generator builds a map a bit larger. In JSON the whole generated map is exported, in PNG - only its visible (requested) part. The point with coordinates (0, 0) in the JSON file will be at the centre of the PNG image.
In MFCG, it's more complicated because the generator exports a map (in PNG and SVG) in such a way that the content (buildings) fills the whole image more or less.