Video see-through still undershoots the eye in dynamic range and resolution; highlights and shadows both lose

What it is

Read a plaque indoors against a window and the highlight blows to white while the letters in shadow vanish. Video-see-through sensors and panels undershoot the retina in both dynamic range and angular resolution. Highlights clip, shadows drown in quantization noise, fine structure disappears on the sampling grid. Both ends drop at once: the dual loss of pass-through dynamic range and resolution, not a vague “slightly worse picture.”

The eye can hold cloud texture and letters under trees across more than a dozen stops. Current passthrough chains cannot.

Why it happens

A silicon well and its read noise give roughly ten to fourteen usable stops; the eye, with pupil and local adaptation, spans more. The display cuts again: peak luminance cannot hold the window, black level cannot hold the corner. Tone mapping keeps the midtones by crushing both ends. Resolution is a separate shortfall: camera pixels over the field of view often undersample the fovea, then compression, sharpening, and the display grid smear the strokes on a nameplate first.

The two shortfalls couple. Drop exposure to save highlights and shadow SNR collapses, wrecking an already thin resolution; raise exposure to save shadows and lamps become structureless blobs, taking edge detection and letter recognition with them. Passthrough is not a slightly blurred reality. It is a picture with both ends truncated and the middle folded.

Studying it

HDR-VDP-style visible-difference prediction plus contrast matching: the same real scene, compared among the naked eye, a passthrough headset, and several tone maps. The target set must include a filament or outdoor cloud in the highlight and small type in the dark.

Conditions: scene luminance span, exposure policy, display peak, camera pixels per degree. Scores: countable highlight features, letter identification in shadow, preference rank, reports that “the window became a white hole.”

Uniform office light hides the whole problem. A valid test puts a window, a lamp, and a dark corner in one frame.

Where it stops holding

Windowless rooms and evenly lit galleries can ignore the gap; people will say it is clear enough. Night roads, backlit lobbies, welding arcs turn the gap into a functional failure. Optical see-through rides real photons and tracks the eye’s own range — its problem is additive contrast, not HDR clipping. Reduced contrast sensitivity from disease or age makes the passthrough loss noticeable sooner. If the task text already lives on the virtual layer, whether the real plaque is readable is no longer in this acceptance test.

Applying it

• Do not put critical reading in a real highlight or a dead black “through passthrough.” Promote the words to a virtual label, or have the wearer turn away from the window.
• Give passthrough a lockable exposure so auto-exposure does not pump between window and wall and repeatedly wreck the midtones.
• In backlight, night, and hard-lamp scenes, drop tasks that depend on real texture (street signs, gauges) and move them onto a high-contrast virtual layer.
• How to check: one window and one shadowed plaque, photographed together in passthrough. If outdoor structure is gone and the plaque letters cannot be read, the shortfall is on the task path. Passing on an office wall does not count.

Related

• Same group：N5.08.1 In optical see-through, virtual content can only add light and cannot truly occlude real objects · 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 · N5.08.5 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
• Search terms：pass-through dynamic range · HDR-VDP · see-through resolution