Stereopsis (2D plan): Disparity and the Horopter

Top-down map of binocular disparity: drag the fixation point and the probe, see the Vieth-Mueller circle, fused and double-image regions and both retinal images.

Switch to 3D

Presets

Playback

Eyes and acuity

Probe target

Layers

Model check

The science

Geometry. The eyes sit at (±I/2, 0). A point P subtends the vergence angle γ(P) = αL − αR at the two eyes, where α are the azimuths of P from each eye. The relative binocular disparity of P with respect to the fixated point F is η = γ(P) − γ(F); positive (crossed) disparity means P is nearer than the horopter.

Horopter. By the inscribed-angle theorem all points with the same γ lie on one circle through both eyes: the geometric Vieth–Müller circle, radius R = I / (2 sin γF). The measured empirical horopter is flatter (Hering–Hillebrand deviation); this lab shows the geometric one.

Depth resolution. For a midline target γ = 2 atan(I / 2z), so a step Δz changes it by about IΔz / z². A stereoacuity δ (about 20 arcsec for good observers) gives Δzmin ≈ δ z² / I, which grows with the square of distance; stereopsis fails where Δzmin = z, i.e. z ≈ I / δ (about 650 m for 20 arcsec).

Panum's area. Only disparities below the fusion limit (about 10 arcmin at the fovea, more in the periphery) are fused into one image; larger ones produce diplopia. The model uses a single constant limit and ignores eccentricity dependence, vertical disparity, torsion and the real distribution of retinal correspondence.

Checks in the page. (1) vergence is constant on the circle; (2) two independent formulas for γ agree; (3) the exact depth step reproduces the stereoacuity angle and the quadratic approximation. Real stereopsis also depends on cortical processing, luminance, texture and monocular cues, which are not modelled. Educational model, not a clinical test.

Disparity vs distance
Smallest resolvable depth step
Retinal images of both eyes