Colour Corrected Smoothest Reflectance Estimation

Mayne, Violet ORCID: https://orcid.org/0009-0003-5192-7586 and Finlayson, Graham D. (2025) Colour Corrected Smoothest Reflectance Estimation. In: Colour Corrected Smoothest Reflectance Estimation. Journal of Physics: Conference Series, 3128 (1). UNSPECIFIED, pp. 1-8.

[thumbnail of Mayne_2025_J._Phys.__Conf._Ser._3128_012012]
Preview
PDF (Mayne_2025_J._Phys.__Conf._Ser._3128_012012)
Available under License Creative Commons Attribution.

Download (737kB) | Preview

Abstract

The same colour response can result from many different spectral reflectances: there are many metamers. This said, there is interest in finding a simple computational method to recover a single ‘good’ metamer for a given sensor and lighting condition. One such method - smoothest reflectance estimation - returns the reflectance whose sum of squared derivatives is minimum over all possible metamers. A weakness in the smoothest reflectance methodology is that different smoothest reflectances are recovered under different lights. In this paper, we try and mitigate this problem by using a colour correction preprocessing step. Simply, we colour correct - by regression - colour responses for a given acquisition light to a fixed reference D65 illumination. We then recover the smoothest reflectance with respect to D65. Experiments show that, on a cross-validation basis, adding the colour correction step can reduce the variability in recovered smoothest reflectances by up to 50%.

Item Type: Book Section
Uncontrolled Keywords: das to be completed ,/dt/atira/pure/keyword/DAStatement/das_to_be_completed
Faculty \ School: Faculty of Science > School of Computing Sciences
UEA Research Groups: Faculty of Science > Research Groups > Colour and Imaging Lab
Related URLs:
Depositing User: LivePure Connector
Date Deposited: 15 Oct 2026 15:18
Last Modified: 15 Oct 2026 15:18
URI: https://ueaeprints.uea.ac.uk/id/eprint/104810
DOI: 10.1088/1742-6596/3128/1/012012

Downloads

Downloads per month over past year

Actions (login required)

View Item View Item