Taking a photo and making it vector is tough, you will find the results of programs that do this automatically are usually unacceptible unless it's a very simple picture with only a few solid shapes (e.g. a wine bottle against a plain background is much easier than a human standing in front of a tree). It's best done by hand with someone with a talent for this sort of thing (you'll find such people at design firms, sign shops, etc).
If you can't find someone willing to do this, or the cost is too high... but you're good with photoshop or gimp, then you can prepare the photo in a way that makes it easier to vectorize, and then vectorize it yourself.
Your goal is to reduce the number of colors and shapes, and get rid of fine details and complex bits. It depends on the final output though, is this metal printing the sort of thing where the final result is everything is cut out of the metal? Does it permit different depths of cutting? If the machine can only cut one depth, then basically there can be no shading... just solid shapes or outlines. Everything is either cut or not cut. In that case, you would want to make the image just 2 colors, one for cut and one for not cut. Something like the photoshop command image --> adjust --> threshold will do this. You could also trace the image in illustrator if you're good with the pen tool.
If you can somehow do multiple depths or colors, you need to find out from the firm exactly how many are possible. Say it's possible to cut 4 different depths... in photoshop you might do something like this: image --> mode --> indexed color --> number of colors (4). Then in addition I'd do filters like dust & scratches or median to clear out tiny details like hair or leaves while leaving larger shapes intact (unless the firm says their machine can handle small details).
