RIT-Dupont数据集，罗彻斯特-杜邦数据集
During the 1980s, a large-scale experiment was carried out at Rochester Institute of Technology‘s Munsell Color Science Laboratory.The goal was to sample CIELAB space in representative object-color regions and in each region, determine CIELAB’s lack of effectiveness in pre dicting supra-threshold visual color tolerances.
DuPont prepared over 1000 samples, high-gloss sprayed alumi num panels using a solid-coat automotive acrylic lacquer. The samples were designed to sample CIELAB in specific univariate directions. The panels were cut and fashioned into color-difference pairs at RIT. The color-difference magnitudes were designed to enable the use of probit analysis, a univariate psychophysical technique that esti mates a response that is equivalent to a reference response when using the method of constant stimuli.
In this case, the CIELAB Euclidean distance along a specific direction from a standard was determined that had the same perceivedtotal color difference to a near-neutral color-differ ence (ancor) pair of 1.02 DE*ab, referred to as the T50 (tolerance at 50% probability).
The visual experiments were performed in two phases, I and II, corresponding to two M.S. theses by Snyder and Reniff,respectively. A third M.S. thesis by Balonon-Rosen analyzed the visual data and improved its monotonicity by using a smoothing filter on the individual observer data.
Several color cen ters and directions were evaluated in both phases to verify data pooling. In addition, several vector directions from Phase I that had large uncertainty were repeated in Phase II.
The data were published as a set of 156 color-differ ence tolerances:
19 color centers with between seven and nine directions about each center and included the standard’s CIELAB coordinates, the vector direction, the T50, and the uncertainty about the T50.
The RIT-DuPont dataset has been used extensively for formula development and testing since its inception during the 1980’s, for example, in the development ofCIE94 and CIEDE2000.The dataset was published a 156 color-tolerances, T50, along specific vector directions about 19 color centers. Probit analysis was used to transform judgments of 958 color-difference pairs by 50 observers to these 156 tolerances. For most statistical significance testing, the number of samples determines the confidence limits.
图来源：Uniform_Color_Spaces_Based_on_CIECAM02_and_IPT_Col.pdf
