{
  "title": "Rectified stereo depth and two distinct uncertainty descriptions",
  "status": "Ideal analytical model, no measured cameras or human/clinical performance",
  "rows": {
    "stereo": 573,
    "boundary_cases": 8
  },
  "parameters": {
    "focal_length_px": 800,
    "baselines_m": [
      0.05,
      0.1,
      0.2
    ],
    "depth_m": [
      1,
      20
    ],
    "depth_step_m": 0.1,
    "disparity_standard_uncertainty_px": 0.5,
    "separate_deterministic_disparity_bound_px": 0.5
  },
  "equations": {
    "disparity": "d=f*B/Z",
    "depth": "Z=f*B/d for positive B,d",
    "local_standard_uncertainty": "sigmaZ=f*B*sigma_d/d\u00b2",
    "bounded_interval": "[fB/(d+delta),fB/(d-delta)] when d>delta; upper unbounded for 0<d<=delta"
  },
  "checks": {
    "all_pass": true,
    "finite_difference_relative_error_max": 1.1081580097993537e-10,
    "inverse_geometry_all_rows": true,
    "interval_endpoint_inversions": true,
    "zero_baseline_and_disparity_noestimate": true,
    "unbounded_upper_cases": true,
    "zero_sigma_and_delta": true,
    "invalid_inputs_rejected": 6
  },
  "reference_example": {
    "depth_m": 2.0,
    "local_standard_uncertainty_m": 0.025,
    "bounded_lower_m": 1.9753086419753085,
    "bounded_upper_m": 2.0253164556962027,
    "upper_unbounded": false,
    "estimate_available": true
  },
  "limitations": [
    "Depth is axial optical-axis depth, not Euclidean camera-to-point distance.",
    "Parallel rectified ideal pinhole cameras with equal known focal length, matched principal points and correct correspondence; no calibration errors or distortion.",
    "Standard uncertainty propagation is first order and conditional on exact f,B. It is not a confidence interval and becomes unreliable for large sigma_d/d.",
    "The separately chosen bounded error is deterministic and not an interpretation of the standard deviation.",
    "No matching outliers, textureless/repeated patterns, occlusion, specular response, motion or drift is modeled.",
    "Wider baseline may impair overlap and correspondence; the ideal formula does not imply universally better reconstruction.",
    "No display cross-talk, tracking latency, visual comfort or tissue-imaging capability is established."
  ]
}