In optical see-through, virtual content can only add light and cannot truly occlude real objects

What it is

Park a “solid” virtual panel in front of a window and the window still glows. An optical see-through combiner can only add display photons to the real light already reaching the eye; it cannot subtract them. The virtual layer is therefore additive optical see-through: it can brighten, tint, and outline, but it cannot cut a true dark patch on the retina that hides the wall, lamp, or face behind it.

Painting a pixel black here means adding nothing. Black is not a mask. It is transparency.

Why it happens

Waveguides and half-mirrors are passive adders. Scene light crosses the optic to the pupil; the microdisplay dumps its own light on top. What the eye gets is the sum of two irradiances. True opacity would require blocking or cancelling the real beam in those directions — a spatial light modulator most production OST headsets do not have. So even a “solid” virtual object still leaks high-contrast edges from behind.

Depth cues arrive incomplete. A real object hiding a virtual one can still be approximated with environment depth; a virtual object hiding a real one cannot. Stereo and motion parallax say “this slab is in front,” while the texture of the wall keeps arriving. The contradiction fails first at near distances and on high-contrast backgrounds.

Studying it

Brightness-matching and occlusion-judgment tasks keep the same virtual geometry and compare additive OST, occlusion-shutter OST prototypes, and video compositing on whether people report a block and on the matched equivalent opacity.

Manipulations: background luminance, peak virtual luminance, occlusion mask on or off. Measures: occlusion accuracy, equivalent opacity, rated solidity, time to resolve the figure on a high-contrast ground.

Labs often use a uniform gray light box. Real windows, lamps, and glossy desks raise transmitted light by an order of magnitude; a panel that “looks solid” indoors turns into fog outside.

Where it stops holding

Research OST with spatial light modulators or LCD occlusion layers can subtract in the optical path; those results do not transfer to production waveguides. In a very dark room with the virtual layer at device peak, additive contrast is briefly enough and people will accept “it blocked it” — daylight reverses the judgment. Monocular OST adds light to one eye while the other still sees the full scene; binocular rivalry further dismantles the occlusion illusion. Tasks that only annotate and point, and never need to hide what sits behind, barely feel this limit.

Applying it

• If the task must hide real clutter, protect privacy, or present a solid panel, do not bet on ordinary optical see-through. Switch to a digitizing path, or to outlines, emissive strokes, and translucent patches.
• Put critical marks in high-luminance fine structure, not large dark fills — the fill cannot hide anything and only spends display power.
• Near a window or lamp, do not encode front/back as “this slab should hide what is behind.” Offset the layout or use a leader line.
• How to check: place a designed “solid” panel in front of a window and in front of a light box, then doff the headset and compare. If the high-contrast edges behind remain readable, the additive path has broken the occlusion assumption.

Related

• Same group：N5.08.2 Video see-through re-digitizes the real world, so a virtual object can occlude a real one · N5.08.3 Camera position is offset from the eye and needs extra correction · N5.08.4 Video see-through still undershoots the eye in dynamic range and resolution; highlights and shadows both lose · N5.08.5 Optical see-through still shows the real world on power loss; video see-through blinds
• Nearby：N5.01 See-through Modes · N5.02 Virtual–Real Occlusion · N3.06 Transparency and Occlusion
• Search terms：additive optical see-through · optical combiner · occlusion-capable display