Datasets / Cherepkova_2025_Contrast_Preferences

Cherepkova et al. (2025) Individual Contrast Preferences Database (NTNU)

individual contrast preferences for natural images (accessibility-relevant)

contrast-pref

Description

Per-observer preferred contrast adjustments for 500 natural images, gathered by Cherepkova and colleagues at the NTNU Colour and Visual Computing Lab (2025). 22 observers each rated their preferred contrast modification for every image; values are signed deltas (negative = less contrast preferred, positive = more contrast preferred). Useful as a perceptual ground-truth for image-adaptive contrast algorithms (APCA / WCAG-adjacent work), and as a sanity test for any color space that claims a perceptually meaningful lightness curve — preferred contrast should correlate with image L*-distribution properties in a well-calibrated space.

cherepkova_contrast.zip
Contains canonical.csv + meta.yaml + raw/
28.7 MB Download

Notes

  • ★ This is a CONTRAST-PREFERENCE dataset, not a color-difference dataset. Adjacent to image quality assessment and accessibility (APCA/WCAG) research, not classical color-space evaluation.
  • 22 observers, 500 images = 11,000 max responses (some cells empty in source).
  • Reference images (JPEGs at zero-contrast adjustment) are bundled in `raw/full_bundle.zip` if you need to compute image-level features. The canonical.csv only contains observer responses.
  • Stratifies well by image content (faces, landscapes, indoor) — see the original NTNU paper for image categorization.

Sample rows

First 5 rows of canonical.csv:

image_idobserver_idpreferred_contrast_delta
1086583obs_1-0.364379
1086583obs_2-0.068233
1086583obs_40.108555
1086583obs_5-0.068589
1086583obs_6-0.125282

Raw files preserved

FileSize
raw/22_observers_preferred_contrast.xlsx164.7 KB
raw/full_bundle.zip28.5 MB

Source publications

Cherepkova, O. et al. (2025). Individual Contrast Preferences Database. NTNU Colour and Visual Computing Lab. Zenodo DOI 10.5281/zenodo.17430959