The font files themselves are copyrighted. The underlying typeface --- the shapes of the letters themselves --- are not, so you could I suppose draw your own Helvetica. But you wouldn't be able to call it "Helvetica", because that's a trademark.
The only thing missing would be hinting, but on retina displays and modern laser printers that's much less important than it used to be -- and you can always implement automatic hinting. And it's easy to extract kerning pairs as well.
I'm not saying this would be good for font creators or society. I'm just wondering why it hasn't become a common thing, when it doesn't seem like it's actually illegal.
(1) It's a lot of work.
(2) You'd get sued anyways and then have to explain to a jury the distinction between your vectorized raster and the original vectors --- I agree that a graphical interpretation of, effectively, a photograph of a curve is not the same thing as that curve, but it's a subtle point.
(3) Most importantly: it just doesn't matter enough. Universally "important" fonts (Helvetica, say) have widely-used liberally-licensed alternatives, but if you want Hoefler Whitney, you want the real Hoefler Whitney, for the same non-pragmatic reason that you'd want a real pair of D&G Daymasters.
But a computer font (as opposed to the typeface it represents) is a specialized program, ie it has logic as well as data. Specifically, kerning tables and the conditions under which to use ligatures would be lost by the approach you're proposing. But I'm not a typographer or font expert and I suspect other things also. Anyway that doesn't make it completely untenable, just not a complete & automatic process.
It is similar in concept to how recipes are handled by intellectual property law. A list of ingredients isn't protected, nor is a specific result. But a list plus instructions is. You can copy a recipe by applying a different set of instructions to the same list of ingredients, resulting in an identical dish. In this the type characters are the ingredients, but you still need to provide a set of instructions to combine them into a font. You can provide ones that create an identical result, but you can't simply copy them from the original font.
But for 99% of fonts it's going to work fine. Kerning tables should be easy to recreate simply by rendering all pairs of characters and measuring the resulting width. And there are only a handful of common ligatures (ff, fi, etc.).
But absolutely, there's a small percentage of fonts with more advanced features that would be harder/prohibitive to reproduce this way.
If the results were bad then, I can't help but think they'd be far better now. I have to assume automatic vectorization is now much improved since the 90's.
[0]: https://www.jdsupra.com/legalnews/my-word-design-patents-on-...