Optical see-through keeps real light but contrast is limited

What it is

Look through a waveguide at a sunlit window and the virtual label is still there, but it has gone watermark-thin. Photons from the room were never rewritten; they now compete with a layer whose brightness has a hard cap. That path is optical see-through: the eye still receives ambient light, and a combiner adds the virtual image. Keeping real light means contrast is pinned by scene illuminance — the hologram can only add a slice of luminance on top of whatever is already arriving.

The question is not who sits in front in depth. It is whether the real world’s brightness still lands on the retina.

Why it happens

Contrast at the eye is roughly (L_v / (L_v + L_r)): the numerator is the virtual luminance the display can deliver, the denominator still includes an unblockable real term. Combiner throughput and diffraction efficiency tax the same microdisplay again at the eyepiece. Two facts then hold together: in a dark room the label can look solid; raise the lux and the same panel washes toward transparency. People read the greying as “not stuck to the wall” or “broken.” It is a photometric budget, not a tracking loss.

A second constraint is adaptation. Walking from a dim interior to a window shrinks the pupil and desensitizes the eye to the added light, so virtual highlights lose relatively more. Optical see-through has no “reality gain” knob. It can add light; it cannot dim the world to make room for the hologram.

Studying it

Place a photometer at eye position, measure peak virtual luminance at the combiner exit, measure ambient in the same seat, and compute usable contrast. Then step illuminance (dark room, office, window, outdoor) against one set of holographic labels. Azuma’s early optical-see-through surveys treat “the real world still arrives as photons” as an architectural premise; later OST headset evaluations treat lux as an independent variable, not a device constant.

Independent variables: ambient illuminance, combiner transmittance, peak luminance and area of the virtual content. Dependent variables: character recognition accuracy, rated solidity, the share of trials in which the label is still judged as part of the scene.

In interface work this decides whether an experience can promise readability in its target venue — shop floor, gallery, street — rather than only in a demo bay.

Where it stops holding

When the target environment stays below office lighting, contrast is rarely the first bottleneck; mass, field of view, and vergence conflict show up sooner. Micro-LED and higher-efficiency waveguides are lifting peak luminance, so outdoor usability is improving — do not freeze “indoor-only” as a property of the architecture. In mines or night work the added light can be too bright and bleach dark adaptation. Monocular OST leaves one eye without the virtual layer, so contrast judgments fight across the two eyes.

Applying it

• Budget contrast for the brightest intended venue. Photograph the same label at a gallery window and under shop-floor lights; if type goes grey, shrink area, add a stroke, or switch to a high-luminance color — do not stack more translucent panels.
• Do not buy readability with a dark backing plate. Optical see-through cannot add negative light; a dark plate on a bright scene is just a fainter film.
• Before a high-lux stretch, give an actionable prompt (step toward a wall, turn your back to the window). The prompt must name something the wearer can actually do.
• How to check: read the same holographic word in a dark room, an office, and at a window. A sharp drop in any bin means the contrast budget missed that venue.

Related

• Same group：N5.01.2 Video see-through is controllable but adds latency and image loss · N5.01.3 See-through mode sets the ceiling on real–virtual fusion
• Nearby：N5.08 Video See-through and Optical See-through · N5.02 Virtual–Real Occlusion
• Search terms：optical see-through · combiner contrast · ambient illuminance