Hong et al. (2025) isoluminant JND ellipsoid grid (Hong, Bouhassira, Chow, Sanders, Shvartsman, Guan, Williams & Brainard)
modern dense JND ellipsoid grid (Wishart Process Psychophysical Model)
Description
Dense per-subject JND ellipsoid field fitted via the Wishart Process Psychophysical Model from Hong et al. (eLife 2025). Eight color-normal observers completed ~12,000 trials each (mix of adaptive AEPsych + MOCS replications). The Wishart Process produces a SMOOTH 2-D field of 2x2 covariance matrices Σ(x, y) across the isoluminant plane (DKL-derived 2DW coordinates, range [-0.85, 0.85]). Each subject's fit is a 103×103 grid of Σ, providing a JND ellipsoid at every point — replacing MacAdam 1942's 25 ellipses with 10,609 per observer. This is the modern direct descendant of MacAdam 1942: fixes both of MacAdam's chief limitations — single observer + sparse sampling — and uses contemporary adaptive psychophysics + Bayesian smoothing.
Notes
- ★ Each row gives a fitted ellipsoid as both the raw 2x2 covariance (sigma_xx, sigma_xy, sigma_yy) and the derived ellipse semi-axes (a, b) plus rotation theta_deg. Both representations are equivalent.
- Grid coordinates (x, y) are in DKL-derived 2DW units, NOT CIELAB. Use the calibration matrices in raw/ (M_2DWToRGB, M_2DWToDKLPlane) to map to RGB or DKL.
- Wishart Process smoothing means adjacent grid points are not independent. Don't treat the 84,872 ellipses as 84,872 independent measurements; effective DOF per subject ≈ ~100.
- Bootstrap iterations (120 per grid point) are NOT included in this canonical export to keep size reasonable. The original Zenodo CSV has all 120 bootstraps if you need confidence intervals.
- ★ This is the gold-standard JND dataset published in the project's diagnostic era; supersedes MacAdam 1942 for isoluminant JND evaluation. MacAdam 1942 remains in the pool for historical comparison only.
Sample rows
First 5 rows of canonical.csv:
| ellipse_id | subject | x_c | y_c | sigma_xx | sigma_xy | sigma_yy | a | b | theta_deg | illuminant |
|---|---|---|---|---|---|---|---|---|---|---|
| hong2025_sub1_000000 | sub1 | -0.85 | -0.85 | 0.0012195 | 0.00056786 | 0.00105287 | 4.135365e-02 | 2.371172e-02 | 40.827 | calibrated display |
| hong2025_sub1_000001 | sub1 | -0.83333333 | -0.85 | 0.00122452 | 0.00056427 | 0.00102245 | 4.119137e-02 | 2.345722e-02 | 39.924 | calibrated display |
| hong2025_sub1_000002 | sub1 | -0.81666667 | -0.85 | 0.00122787 | 0.00056022 | 0.00099445 | 4.102935e-02 | 2.321448e-02 | 39.116 | calibrated display |
| hong2025_sub1_000003 | sub1 | -0.8 | -0.85 | 0.00122956 | 0.0005557 | 0.00096875 | 4.086503e-02 | 2.298606e-02 | 38.397 | calibrated display |
| hong2025_sub1_000004 | sub1 | -0.78333333 | -0.85 | 0.0012296 | 0.00055069 | 0.00094526 | 4.069614e-02 | 2.277464e-02 | 37.762 | calibrated display |
Raw files preserved
| File | Size |
|---|---|
raw/ComputeTransformMatrices.m | 2.8 KB |
raw/M_2DWToDKLPlane_DELL_02242025_copy.csv | 73 B |
raw/M_2DWToRGB_DELL_02242025_copy.csv | 55 B |
raw/M_DKLPlaneTo2DW_DELL_02242025_copy.csv | 71 B |
raw/M_LMSToXYZ1931_DELL_02242025_copy.csv | 156 B |
raw/M_RGBTo2DW_DELL_02242025_copy.csv | 103 B |
raw/M_RGBToXYZ1931_DELL_02242025_copy.csv | 152 B |
raw/M_XYZ1931ToLMS_DELL_02242025_copy.csv | 188 B |
raw/M_XYZ1931ToRGB_DELL_02242025_copy.csv | 183 B |
raw/README.pdf | 324.4 KB |
raw/TransformMatrices_README.md | 4.6 KB |
raw/single_2d_colorDiscrimination_EAVC_4strats_new.ini | 863 B |
raw/single_4d_colorDiscrimination_EAVC_4strats_new.ini | 1.0 KB |