Overlays occlude the real environment
Aliases: HUD occlusion · fill factor · overlay coverage
What it is
A translucent panel over a real doorknob subtracts that patch from the view. Overlay occlusion is literal coverage: every degree of visual angle the overlay occupies is a slice of the real environment gone. It happens before “is this information useful” — a piece of the room is already unseen, whether or not the overlay was worth it.
Density here is first the product of area and opacity, not a feeling about how many items there are.
Why it happens
In optical see-through, the overlay is added light: reality is still there, but contrast is crushed and detail goes from resolvable to a blob. In video see-through, the compositor replaces real pixels with overlay pixels, and the covered patch need not enter the eye at all. The form of the loss differs; the shared result is that usable visual angle on the real environment falls with overlay fill factor. What people use to notice a door opening, a step in the floor, someone approaching, is that angle. Larger, more solid, parked on the fovea — the loss is less reversible.
Strokes, wireframes, and very low opacity buy less area inside the same mechanism. They are not “no occlusion.” They push fill factor down.
Studying it
Make overlay coverage the independent variable: paint panels of different area and opacity on the see-through view, and have people detect a change in the real environment (a door, an object added to a table). Aviation HUD work has long used “how much of the outside do the symbols cover” to predict detection. Dependent measures: detection rate of the real change, time to detect, and which patch they later mark as having been hidden.
Independent variables: overlay fill factor, opacity, whether the overlay sits on the fovea. Dependent variables: hit rate on real changes, whether misses fall inside the overlay’s projection.
A static lab background underestimates the loss — in the field the environment changes on its own, and the hidden patch is exactly where the change happens.
Where it stops holding
A fully virtual level with see-through off has no real environment to cover, so the claim does not apply. Hairlines, a point cursor, a nearly clear outline can sit at a fill factor the task does not feel. When the overlay lands in a direction that already holds no information (a ceiling corner, an empty wall), occlusion still happens and environmental loss is near zero. A small label on a far landscape costs less than the same label on a near work surface.
Applying it
- Give overlays a fill-factor budget. Default panels should not carpet the middle of the view; prefer a wireframe to a solid plate.
- Move large solids off the real surface being looked at. If they must overlap, drop opacity so the surface contour remains.
- How to check: with overlays on, make one real change (close a door, set a cup on the table) and see whether people notice without being asked. A miss whose change sits inside a panel’s projection is the overlay having covered that slice of the room. Count overlay pixels on a screenshot as the live budget number.