Don't you control that by doing text recognition and always placing a frame behind? Let the user customize the color schemes in question (bright white + black text, or off white + off black etc).
If the background is a blue wall in your living room, you place a white framing layer behind the text you're looking at. If the real wall is black, you do the same thing. Makes no difference what color the wall is then.
Wall -> Frame Layer -> Text
Text recognition should be among their easiest chores (which is to say it's still not easy, it's on the lower level of difficulty in what they're trying to do).
The issue is that if you have a transparent display for AR, light from the background goes through it. There's no way to just "put a white framing layer behind it" because the light the display puts out is being added on top of whatever light is coming through from behind it. This is what happens with Microsoft's HoloLens headsets.
Magic Leap is claiming to have occlusion of background objects working, but hasn't really explained the mechanism. It sounds like it's some sort of "light field" trickery where they let the light through, but your brain knows there's a virtual object in front of it and mentally processes it out. Cool if it works, but I'll need to see this to believe it.
If the background is a blue wall in your living room, you place a white framing layer behind the text you're looking at. If the real wall is black, you do the same thing. Makes no difference what color the wall is then.
Wall -> Frame Layer -> Text
Text recognition should be among their easiest chores (which is to say it's still not easy, it's on the lower level of difficulty in what they're trying to do).