Supplementary material for “Impact of the adapted white point and the cultural background on memory color assessments”
Measured reflectance spectra of test objects and Gaussian fit function parameters describing the respective memory color assessments of Chinese and German observers at 3200 K and 5600 K adaptation conditions.
With their inherent ability of serving as an internal reference, memory colors provide a very powerful concept in the evaluation of color rendering properties of white light sources with respect to visual appreciation. Recent results for example suggest fairly good correlations between memory-based color quality metrics and the observers' general color preferences. However, due to limited conceptional design of the psychophysical experiments used to derive these kind of colorimetric measures, it remains doubtful whether these models are capable of adequately predicting the impact of realistic viewing and adaptation conditions on the assessment of memory colors over a wide range of different lighting situations. In addition, inter-cultural effects might play a crucial role, which, so far, has been considered only rudimentarily in the context of memory colors. For these reasons, the impact of both the adapted white point and the observers' cultural background on memory color assessments should be investigated in order to establish a better understanding of these dependencies and their interactions. For this purpose, the color appearance rating results of Chinese and German observers were collected for a selection of twelve different familiar test objects assessed under two different adaptation conditions at 3200 K and 5600 K, respectively. The measured reflectance spectra of these twelve test objects will be presented and the corresponding raw data as well as the Gaussian fit parameters for modelling the color appearance ratings will be provided for download.
The twelve familiar test objects were selected by evaluating an online survey (Link), which was conducted in advance of preparing the final experiments. Based on this survey, the following twelve test objects were chosen: banana, green salad, broccoli, blue jeans, blueberry, red cabbage, red rose, carrot, butternut squash, and concrete flowerplot. In addition two different kinds of human complexion, i.e., Asian and Caucasian skin, were added. Their corresponding spectral reflectance curves were calculated from spectral radiance measurements of the objects' surfaces, which were homogeneously illuminated with a temporally stable halogen light source. These measurements were conducted using a Konica Minolta CS-2000 spectroradiometer in combination with a Spectralon® reflectance standard. For each test object, several measurements of characteristic surface points were averaged to come up with the finally used spectral reflectance curves shown in Fig. 1. The respective raw data for plotting these curves can be downloaded here (Link).
Once the test objects were defined, a specially designed experimental setup was used for collecting color appearance rating data. It consists of a four-channel LED panel to set the ambient illumination and an LCD projector used for shifting the test objects’ chromaticities while keeping the observers’ state of chromatic adaptation constant. For each test object, approximately 65 different color settings could be realized by this setup. During the following experiments, observers were asked to rate for each of these settings the perceived color appearance of the respective test object according to their preference of how they thought it should actually look like in reality adopting a semi-semantic five-level rating scale. These memory-related color appearance ratings collected for both cultural observer groups and the two different adaptation conditions were subsequently modelled in CIECAM02-UCS by fitting bivariate Gaussian similarity distributions of the form
to the corresponding average observer rating data obtained for the different chromaticities of the i„th“ test object. Here, the functions’ centroids μ_i=(a_(i,3),a_(i,4) )^T give the most likely representations of the test objects’ memory colors and the covariance matrices defined by the parameters a_(i,5) to a_(i,7) determine the size, shape, and orientation of the similarity distributions S_i(x_i ). Hence, the characteristics of the memory color assessments of Chinese and German observers at the two different adaptation conditions for each of the twelve familiar test objects are fully described by four different sets of the fitting parameters a_(i,3) to a_(i,7). These sets as well as the corresponding goodness-of-fit measures can be downloaded here (Link).