Abstract
Since the 1980's, the RIT-DuPont dataset has been used extensively for color-difference formula development and testing. In the past, the dataset was published as 156 color median tolerances along specific vector directions about 19 color centers. The median tolerance, referred to as T50, is the estimated absolute magnitude of the color difference at which 50% of the population will accept or reject a pair relative to an anchor pair. 1D Probit analysis, a univariate statistical technique that locates a median threshold from a binary response data, was used to transform judgments of 958 color-difference pairs by 50 observers to 156 T50 tolerances. In order to recover the losses of colorimetric values of RIT-DuPont data, a symmetric sampling method resulting in no significant differences in the color difference calculations was used. In addition, a method based on Gaussian function was devised to assign a visual uncertainty weighting to each of these pairs using the method of maximum likelihood. The weighted dataset was found to be equivalent to the T50 tolerances. Also a visual experiment was performed to generate suprathreshold tolerances sampled in the IPT chroma direction thereby extending the RIT-DuPont dataset. Forty color centers including 20 hue angles at two different lightness and chroma levels for each hue angle were evaluated for chroma discrimination. Fifty observers participated and a total of 14,000 visual observations were made. Probit analysis was also used to determine the chroma discrimination threshold T50_[delta]C, color difference threshold T50_[delta]E, and fiducial limits for each color center in IPT space. The chroma tolerances increase with the color center's chroma. After transforming from 2º to 10º standard observer, this new chroma dataset was used to optimize a new color-difference formula in CIELAB, the formula indicating the need for a nonlinear correction to chroma.
Library of Congress Subject Headings
Color vision--Mathematical models; Colorimetry--Mathematical models; Probits
Publication Date
11-1-2010
Document Type
Thesis
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Advisor
Berns, Roy
Recommended Citation
Hou, Bingxin, "Extending the RIT-DuPont suprathreshold data set: Weighted individual discrimination pair data and new chroma dependency visual data" (2010). Thesis. Rochester Institute of Technology. Accessed from
https://repository.rit.edu/theses/4568
Campus
RIT – Main Campus
Comments
Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: QP483 .H68 2010