Option three can be used in two different ways. You can have a constant color overlay with variable transparency, or a constant transparency overlay using variable color.
Variable transparency would work well at separating foreground items from background while retaining the color of the original, but would give extremely low depth resolution. By this I mean that your 'background' distance would be fully opaque, and get more transparent as the item gets closer.
The problem is that we can only really detect, with any kind of accuracy, between 3 and 6 levels of transparency. This means that you will get general depth information, but the human eye cannot discern between subtle levels of transparency the same way it can discern subtle shades of gray. So the perceptual resolution is far lower. That makes this method excellent for integrating the data into one view, but poor if subtle differences need to be seen.
A constant transparency with varying colors, as in a relief map overlay, would often allow users to perceive depth more accurately (all the red items are at the same depth, etc), but would obscure the original video more. It would be harder to tell what the original colors looked like, obscured as they were by a veritable rainbow. I have not tested it, but I would not be surprised if it took more training to translate colors into depth; this may work better for a professional, but not as good for a beginner tool or demonstration.
So it depends on what you need. Both visualizations would reasonably well depending on what is important to you.
