Hannah E. Smithson

The paper shows that human observers can improve color discrimination by tilting surfaces to sample directional lighting, rather than relying on uniform illumination.

The paper investigates human color constancy in complex lighting environments, finding it remains robust despite misleading global color cues.

This paper investigates the relationship between detection and localization in human vision, showing that localization is optimized by specific eye movement scales while detection remains robust.

The paper shows observers use highlight geometry, not just brightness, to estimate illuminant color from specular highlights in rendered scenes.

This paper shows that tiny neural networks can accurately replicate human gloss perception judgments better than larger networks estimating physical reflectance.

This paper models fixational eye movements in a theoretical framework to analyze their impact on visual information and adaptation in human vision.

The paper shows that directional variability in real-world lighting can resolve most cases of illuminant metamerism, benefiting color-deficient observers more than trichromats.

This study measures how well participants can judge color and gloss of objects under different lighting conditions, finding high accuracy in hue but poor performance in chroma, lightness, and gloss.

The paper models color discrimination thresholds under different lighting conditions using chromatic statistics and CNNs, finding that human-response-trained CNNs outperform chromatic statistics models.

The paper shows that low-level image features can reliably predict metallicity perception, even when objects are rotated.

The paper introduces a portable hyperspectral imaging system to capture spectrally and spatially resolved light environments in woodland areas, demonstrating its application in Wytham Woods.

This paper introduces two physiologically relevant metrics to evaluate how well visual displays reproduce photoreceptor signals, using real-world spectra and various illuminants.

The paper captures hyperspectral environmental illumination maps with 33 spectral channels, revealing directional spectral variation in natural environments.

This paper demonstrates how modified trace learning rules enable a neural network model to transform eye-centred to head-centred visual representations in primates, even with monotonic gain modulation.