Cross-View Variance Correlation in Path-Traced Stereo: A Hidden Shortcut in Synthetic Training Data

Path-traced synthetic stereo data underlie a large fraction of modern
disparity-estimation training pipelines. We report a previously
unrecognised property of such data: while the Monte~Carlo (MC) noise
streams of the two cameras are statistically independent, the underlying
\emph{variance fields}---deterministic per-pixel functions of the
rendering integrand---are highly correlated once aligned by the
ground-truth disparity warp. Across 20 scenes rendered with Mitsuba~3,
the warped Pearson correlation reaches $\rho{=}0.754{\pm}0.016$, and
remains essentially invariant ($\rho{=}0.778{\pm}0.001$) over a
$16\times$ range of samples per pixel. The effect is strongest in
Lambertian regions ($\rho{\approx}0.78$) and substantially weaker in
glass ($\rho{\approx}0.30$), as predicted by an integrand decomposition
into view-independent and view-dependent components. Because real
binocular sensors carry independent thermal and shot-noise streams, the
correlation is unique to MC-rendered data: it constitutes a learnable
shortcut for stereo networks and a previously unrecognised mechanism of
the sim-to-real gap that affects the entire image, not only complex
materials.