Article: Kubelka-Munk Theory for Color Image Invariant Properties (2002)

Title: The Kubelka-Munk Theory for Color Image Invariant Properties
Author: Jan-Mark Geusebroek, Theo Gevers, Arnold W.M. Smeulders
Reference: First European Conference on Colour in Graphics, Imaging, and Vision, Springfield, USA, 2002, pp. 463-467
Link: http://dare.uva.nl/... (pdf, 112 Kb)

Abstract:

A fundamental problem in color image processing is the integration of the physical laws of light reflection into image processing results, the problem known as photometric invariance. The derivation of object properties from color images yields the extraction of geometric and photometric invariants from color images. Photometric invariance is to be derived from the physics of reflection. In this paper, we rehearse the results from radiative transfer theory to model the reflection and transmission of light in colored layers. We concentrate on the Kubelka-Munk theory of colored layers, which is posed as a general model for color image formation. The model is used for decades in the painting and printing industries, and is proven to be valid for a wide range of materials. We relate the Kubelka-Munk is proven valid may be inherited to algorithms based on newer models. Furthermore, photometric invariant properties proven for one model are, by using Kubelka-Munk, easily extended to related models.

Article: Digital Imaging of Cultural Heritage

Title: High-Accuracy Digital Imaging of Cultural Heritage without Visual Editing
Authors: Roy S. Berns, Lawrence A. Taplin, Mahdi Nezamabadi, Yonghui Zhao and Yoshio Okumura
Download: http://www.art-si.org/... (pdf, 258 Kb)

Abstract:

At Rochester Institute of Technology, a research program is near completion aimed at benchmarking the quality of direct digital imaging of cultural heritage in American museums, libraries, and similar institutions. The current practice at nearly all institutions surveyed includes visual editing. Digital masters incorporate camera spatial processing, ICCt ype color management including encoding in a large-gamut RGB space, and global and local visual editing. Also at RIT, a research program is underway aimed at developing a high quality digital camera that incorporates spectral imaging. The hypothesis is that when using the new camera system, visual editing is unnecessary, greatly improving workflow efficiency and color accuracy. An experiment was performed to test this hypothesis. The experiment included spectral based imaging of both color targets and small paintings and rendering the spectral images for a colorimetrically characterized computer-controlled LCD display. The targets and paintings were viewed adjacent to the display in a laboratory lit by ceiling-mounted daylight-balanced fluorescent lights. A variety of quantitative comparisons were performed including: reflectance spectrophotometry vs. in-situ spectroradiometry, reflectance spectrophotometry vs. spectral-based imaging, forward and inverse model accuracy of the LCD colorimetric characterization, and in-situ spectroradiometric comparison of targets and paintings compared with their LCD renderings. Using the GretagMacbeth ColorChecker as an independent verification target, average color differences varied between 1.0 and 2.9 E 00. For two paintings, the average accuracy was 4.2 and 5.1 E 00. This level of accuracy exceeded that achieved by museums and libraries, even following global and local image editing, confirming our hypothesis that it is possible to create a digital archive of cultural heritage without the need for visual editing.


In case you would like to share this publication, please acknowledge this source.