John Platts’s answer to Why does a mirror reverse things horizontally but not vertically? includes some nice illustrations and discussion but declares that the front-to-back reversal is not horizontal.
Interesting point. But I think we use the word “horizontal” to refer to any part of a line or plane that is perpendicular the line joining it to the Earth’s centre, regardless of whether or not our view of it actually appears parallel to the horizon. We often test for this property by looking at it from a position where it lines up (left-to-right) with the horizon but two people building a house will generally agree that a beam is horizontal even when they are not looking at it from such a position. (If that were not the case the word would be a lot less useful because the property of a beam being horizontal would depend on the observer and would be lost whenever she changes her orientation.)
So, in my understanding, horizontal lines aren’t necessarily parallel to the horizon in a perspective view, but (if the world was flat) their infinite extensions (would) all terminate on it (though due to the Earth’s curvature that actually happens a tiny bit above what we see as the visible horizon).