No-Reference Error Detection in Color Difference Datasets: Application to Munsell Data
The Munsell dataset is a cornerstone in color science, offering extensive coverage of large color differences across a wide gamut, and serving as a prominent resource for developing and validating color models. The widely used second version, the “Munsell Renotation” dataset, h...
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11048487/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The Munsell dataset is a cornerstone in color science, offering extensive coverage of large color differences across a wide gamut, and serving as a prominent resource for developing and validating color models. The widely used second version, the “Munsell Renotation” dataset, has been instrumental in advancing uniform color spaces (UCS). However, substantial errors in the third version, the “Re-renotation”, compromise its reliability for optimizing color models. In this study, we identify and correct typographical errors in the “Re-renotation” dataset and introduce two semi-automatic methods for error detection. The Global Metric Outlier Detection (GMOD) method applies to datasets with varied structures, while the Smooth Curve Outlier Detection (SCOD) method targets ordered datasets, such as the Munsell dataset. These methods operate without relying on external references, instead assuming that a true color space metric guarantees gradual and smooth perceptual change between colors. Our revised dataset demonstrates significantly improved consistency with established UCS models. The corrections and methods introduced here support the ongoing refinement of psychophysical color difference datasets, with all results and data made available in a machine-readable format. |
|---|---|
| ISSN: | 2169-3536 |