Various color difference metrics were developed for characterizing the perceived color difference between individual color patches. Color difference between palettes containing multiple color patches, however, is critically important in product design and computer graphics. This study aimed to investigate how the perceived color difference between a pair of color palettes containing more than a single color patch is affected by the order and number of color patches in the palette. Two reference color sets were generated and each set had four color palettes containing 1, 4, 9, and 16 color patches that were arranged as 1 × 1, 2 × 2, 3 × 3, and 4 × 4 patterns. Human observers scaled the color differences between a color palette of the reference set and a color palette that had revised colors, or revised orders, or a combination of revised colors and orders compared to the reference palette. The calculated color differences between the two palettes were derived using the Minimum Color Difference Model (MICDM) algorithm proposed in a recent work with different color difference metrics, including CIELAB, CMC, CIE94, and DE2000. It was found that the perceived color differences of pairs of individual color patches were significantly larger than those containing multiple patches, when the calculated color differences were the same. The color differences metrics, except for CIE94, had similar performance when characterizing perceived color differences between color palettes.
The difference or distance between two color palettes is a metric of interest in color science. It allows a quantified examination of a perception that formerly could only be described with adjectives. Quantification of these properties is of great importance. The objective of this research is to obtain the dataset for perceptual colour difference between two color palettes and develop color difference metric(s) to correspond well with the perceptual color difference. The psychophysical experiment was carried out using Magnitude Estimation method. Three different color difference metrics, namely Single Color Difference Model (Modell), Mean Color Difference Model (Model 2), and Minimum Color Difference Model (Model 3), respectively, have been proposed and compared. Data analysis include regression analysis, statistical STRESS analysis, and examination of observer variability using coefficient of variance (CT). The results show that the Minimum Color Difference Model (Model 3) outperformed the other two with a coefficient of determination (R-squared) value of 0.603 and an STRESS value of 20.95. In terms of observer variability, the average intra-observer variability is 17.63 while the average inter-observer variability is 53.73.
In many medical test designs, presence of a color spot can represent existence of a disease. This presence is usually verified by observation. For these tests, knowing the color discrimination threshold is necessary for modifying the indicator spots to have color differences above the threshold. In this work, a psychophysical experiment is used to determine the color difference threshold for a veterinarian test device from IDEXX company with blue-green spots. The study was conducted in two phases, In the first phase a preliminary investigation was conducted for the ideal situation, that is having perfect circular spots without any noise or non-uniformity. Method of constant stimuli was used to present designed test images to the observers. The results were analyzed using Probit analysis method. The second phase of the study was performed with objective of studying the effects of noise, imperfect spot shapes, presence of streak and presence of spot color gradient on the color difference threshold between the background and spot colors. The same experimental and analysis method was used in both phases. The results for the ideal situation showed an average discrimination threshold of 1.27 color difference (DE00) for overall data. For the realistic situation, the noisy appearance of the image and imperfectness of the shape of the spots did not affect the threshold when observers were expecting imperfect spots. However, the presence of streak and spot gradient increased the threshold.
To examine the performance of a select group of advanced color difference equations against visual color difference data, we report the development of a combined visual dataset consisting of samples in the CIE low and high chroma blue color centers (NCSU-B1 [1] and NCSU-B2 [2]), a recent set of near black samples (NCSU-BK) [3] and a new dataset around a gray center (L* =50.56, a* =-0.11, b* =0.03), hereafter called NCSU-Gr, using the gray scale method. The new gray dataset consisted of 21 matte painted samples, and the visual difference between each of the samples against the standard was assessed by 35 color normal observers under highly controlled viewing and illumination conditions and using the AATCC gray scales, in three separate sittings, and a total of 2205 assessments were obtained. The performance of two groups of color difference equations consisting of: 1- those based on CIELAB color space and 2- those based on more uniform color spaces/appearance model such as DIN, CIECAM02 and OSA, against the visual dataset was examined for the NCSU-Gr, and also for the combined dataset (NCSU-COM). The results show that CIEDE2000 (2:1:1) exhibits the best performance for the NCSU-Gr dataset in comparison to other equations examined. This confirms that the G term in the CIEDE2000 significantly improves its performance in the near neutral gray region. An examination of the performance of the models against the combined dataset, however, shows that the more uniform color space/appearance models produce better results than models based on CIELAB color space, with CAM02-SCD performing significantly better than other equations except CAM02-UCS.