Premium
Automatic color classification via Munsell system for archaeology
Author(s) -
Milotta Filippo Luigi Maria,
Tanasi Davide,
Stanco Filippo,
Pasquale Stefania,
Stella Giuseppe,
Gueli Anna Maria
Publication year - 2018
Publication title -
color research and application
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.393
H-Index - 62
eISSN - 1520-6378
pISSN - 0361-2317
DOI - 10.1002/col.22277
Subject(s) - rgb color model , color space , artificial intelligence , computer science , computer vision , ground truth , icc profile , lightness , color correction , color model , computer graphics (images) , remote sensing , geology , image (mathematics)
Munsell soil charts are commonly used in archaeology to identify colors of soils and artifacts during excavations. In situ , the procedure is done by visual means and in a very subjective way, resulting in a time consuming and error‐prone approach. To overcome these limitations, an original application has been developed: the automatic recognition of color for archaeology (ARCA). ARCA is a framework thought to be a valuable asset for archaeologists since it may be objective, deterministic, and fast. In this framework, it is possible to convert RGB data to HVC color space. Users may import images in ARCA through a proper desktop application, which also allows operators to sample RGB manually data. Then, HVC notation is automatically estimated and provided to operators within an automatically generated filled report. ARCA moves from Munsell charts, but it has the advantage of providing an objective and affordable color specification system. In this study, we present the results related to data acquired in a controlled lightning environment with a color assessment cabinet on Munsell soil color charts to try to improve the lightness estimation. We focused on two color space conversions: RGB to HVC, and RGB to L * a * b * of CIELAB color space. Color coordinates are obtained through colorimetric measurements. We developed a method with three main phases, and we computed transformation coefficients from observed and ground truth data. The color accuracy of our method is presented in terms of Godlove distance and through CIEDE1976, CIEDE2000, Δ L *, Δ a *, and Δ b * CIE metrics.