Open Access
Add to BibSonomy
Abstract
Provoking high user acceptance in lighting can be a very challenging task and demands suitable tools for properly modelling and predicting the users’ perception of the lit environment. Recently, a new model formalism based on the perceptually relevant attributes of perceived brightness, visual clarity, and color preference has been introduced and successfully applied in some preliminary studies. However, a proof of the model’s applicability from a lighting practitioner’s point of view for realistic lighting scenarios and use-cases is still pending and should be performed as part of this work. For this purpose, results of two dedicated lighting condition rating experiments representing different lighting contexts are reported. It is shown that the model predictions for all three perceptual attributes exhibit excellent linear correlations with the respective subject mean ratings and, in all cases, correctly predict the test light sources’ rank order. These results clearly emphasize the applicability and practical relevance of the model and confirm the suitability of its multi-dimensional approach.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
A fundamental goal of modern interior lighting design is to create lit environments that integratively satisfy the users’ visual, emotional, and biological needs [1–3]. In this context, lighting quality can be understood as a measure of the appropriateness of the lighting conditions to provoke desirable outcomes in the human user [4]. Besides taking into account the non-imaging forming aspects of lighting, which have been extensively studied and discussed in recent years (see e.g. [5–8] for corresponding reviews), it is particularly important to include emotional factors for achieving high user acceptance and visual appreciation. Even though it might result in preferable biological outcomes, a lighting condition that is disliked by the users could never be considered as a good option.
Modelling user preference in the lighting context to be used as an integral part of a sophisticated lighting quality model, however, is a very challenging task. Excluding daylight impact, there are basically three different degrees of freedom for a lighting designer to adapt the lighting conditions in a room to match the users’ emotional needs: i) Light intensity, which goes hand in hand with the perceptual attributes of perceived brightness [9–11] and visual clarity [12–17]; ii) Spectral power distribution (SPD), which determines the perceived white point [18–21] and correlated color temperature (CCT) [22–24] of the illumination and affects the color appearance of objects [25–29]; iii) Spatial light distribution [30–42], which is determined by the individual luminaires constituting the lighting system and their orientation in the room.
As a first attempt to comprehensively model the emotional complexity of lighting, Khanh et al. [43,44], based on a series of recently conducted visual rating experiments, developed a mathematical formalism intended to predict the users’ feelings of perceived brightness (B), visual clarity (VC), color preference (CP), and scene preference (SP) for a given lighting condition. For this purpose, subjects were asked to assess their visual impressions in terms of each of these perceptual attributes for different still-life arrangements presented to them in a real room, while light intensity and the SPD of the illumination were varied systematically. The spatial light distribution was adjusted such that a homogeneous illumination of the still-life arrangements was achieved for all test conditions. Based on the subjects’ ratings, each attribute was subsequently modelled by using reasoned combinations of the corresponding lighting parameters illuminance $E_{\mathrm {v}}$, CCT, and saturation of the illuminated objects $\Delta C^{*}$.
Such being the case, the work of Khanh et al. is among the first to pursue a multi-dimensional approach for modelling emotional aspects of lighting. Existing color quality metrics that are often used to predict color rendering and preference in lighting applications for example lack the possibility of taking into account variable illuminance levels. In a recent study by Kawashima and Ohno [45], in which subjects were asked to rate the appearance of real fruits and vegetables as well as their own skin tone for different levels of saturation at two different illuminances of 100 lx and 1000 lx, respectively, it was found that subjects preferred less saturation at higher compared to lower light levels. Similar findings were reported by Wei et al. [46]. In their work, subjects were invited to assess the color appearance of oil paintings illuminated by four different metameric light spectra provoking distinct levels of object saturation for a fixed CCT of 3000 K at two different illuminances of 20 lx and 500 lx, respectively. Again, it was found that the preferred level of saturation decreased with increasing illuminance. A likely explanation for these empirical results is given by the Hunt effect [47], stating that with increasing light levels, the perceived colorfulness of colored objects increases too. Thus, it can be expected that at higher illuminance levels but constant white point adaptation, less (over-)saturation is required to provoke the same feeling of preference in the observers.
In summary, it can be stated that explicitly considering the impact of illuminance is crucial for any model of lighting quality to be applicable in practice, not only in terms of predicting brightness perception but also with regard to modelling color or scene preference. Preferred illuminance levels, primarily in relation to the former, have extensively been studied in the past. Most of these studies, performed around the 1960s, focused on the context of office lighting [48–56]. From these studies, it can be concluded that user-preferred horizontal illuminance levels are well above the normative minimum requirements of today’s (international) lighting standards. Typically, preferred light levels were found to range from 1300 lx [49] up to even 2500 lx [53]. This tendency for much higher illuminances to be preferred by the users was confirmed by two more recent field studies in office environments with daylight entry conducted by Begemann et al. [57] and Moosmann [58], even though slightly lower values for preferred illuminances ranging between 850 lx and 1400 lx were found in both cases.
The large variance of illuminance levels reported in the literature to be preferred by users or judged as sufficiently bright in combination with the fact that the corresponding experiments were often performed at significantly different CCTs promote that photometric illuminance alone is not sufficient to properly describe the perceived brightness of the lit environment [11]. Indeed, Boynton [59] argued on the level of cone excitations that also short-wavelength components of a light source’s SPD must contribute to its brightness impression, giving an explanation for the observation that light stimuli of equal luminance but different CCTs provoke a distinct brightness perception [60]. Based on previous work by Marks [61], it could be shown that this phenomena particularly depends on the degree of S-cone excitations in such a way that higher CCTs (larger S-cone excitation) are judged brighter than lower CCTs at otherwise equal photometric conditions. Motivated by these findings, Fotios and Levermore [62] developed a brightness perception model that explicitly considers S-cone contributions. Due to its simple form and excellent positive correlation with the more sophisticated brightness perception models known from literature (e.g. [10,63–65]), the Fotios-Levermore model was adopted in the work of Khanh et al. [44], who additionally showed that it can be approximated with sufficient accuracy using a polynomial expression of CCT.
Higher CCTs, however, do not only increase perceived brightness but to some extent also lead to higher ratings of color preference [25,28,29,66–68] and enhance visual clarity [69,70] as defined by the capability of clearly discerning continuous color transitions, fine color shadings on the object surfaces, and subtle contrasts between similarly colored objects [69]. Due to these non-negligible interactions between illuminance, CCT, and object saturation considerably affecting how people perceive the lit environment, addressing a multi-dimensional approach for modelling lighting quality as proposed by Khanh et al. seems to be expedient with regard to practical applications. Indeed, in two recent published studies on the emotional responses to lighting in the museum context [71] and on the color preference of variable lighting conditions in a viewing booth experiment [72], which both investigated the impact of different levels of illuminace and CCT, post hoc analysis performed by the respective authors confirmed in both cases the excellent correlation of Khanh et al.’s model with the collected observers’ preference ratings. Despite these promising results, proving the model’s applicability for realistic lighting scenarios and use-cases from a lighting practitioner’s point of view is still pending and should be performed as part of this work. Two dedicated experiments have been conducted accordingly. Their results will be presented and discussed in this paper.
For this purpose, it is organized as follows. In Sec. 2, the model formalism as well as the experimental setups, the test procedure, and the methods for data analysis are presented. Sec. 3 then summarizes the results of the two experiments including the break-down of statistically significant main and interaction effects. Subsequently, in Sec. 4, corresponding conclusions about model performance are discussed with regard to the model’s practical relevance. Finally, Sec. 5 briefly summarizes the main findings and provides some concluding remarks.
2. Materials and methods
2.1 Model formalism and perceptual attributes
The following sections are intended to give a brief overview of the mathematical formalism developed by Khanh et al. [43,44]. For further details on the background of the empirical modelling (including a discussion of the underlying experiments) and the composition of the individual terms, the interested reader is referred to their original work.
2.1.1 Perceived brightness
The perceived brightness B is predicted by
where the equivalent illuminance $E_{\mathrm {v,eq}}$ is calculated by correcting the illuminance $E_{\mathrm {v}}$ with the Fotios-Levermore S-cone ratio: where $S$ is the integral over wavelength of the light source’s relative SPD weighted by the spectral S-cone sensitivity, while $V$ denotes its $V(\lambda )$-weighted counterpart. Khanh et al. [44] have shown that the term $(S/V)^{0.24}$ can be approximated with sufficient accuracy using a polynomial expression of the form which should be used for the rest of this work.2.1.2 Visual clarity
The expression for calculating visual clarity VC is given by
2.1.3 Color preference
The color preference model prediction CP is obtained from
The additive correction term further accounts for the observers’ white point and CCT preferences [19,22–24] and is modelled as a polynomial of the Fotios-Levermore S-cone ratio. It basically gives negative values for warm-white ($\lessapprox {3600}\,\textrm{K}$) and cool-white ($\gtrapprox {9200}\,\textrm{K}$) light sources and positive values for rather neutral light sources with a relatively broad maximum in the range from approximately 5000 K to 7800 K.
2.2 Experimental setup
In order to validate the applicability of Khanh et al.’s lighting quality model described in the previous section from a practical point of view, two similar experiments had been designed, in which subjects were invited to assess the lighting conditions for two different mock-up scenarios set up in a real-sized, white-painted room ($\mathrm {footprint:\;} 5.5\,\mathrm {m}\,\times \,4.5\,\mathrm {m,}$ $\mathrm {height:\;} 3.3\,\mathrm {m}$). As can be seen from Fig. 1(a), the first mock-up mimicked a typical supermarket display of fruits and vegetables. Care was taken to include a broad variety of object colors and, thus, of different fruits and vegetables, including ripe bananas, red and white cabbage, plums, green, yellow, orange, and red bell peppers, cucumbers, green and red apples, oranges, lemons, carrots, and tomatoes. The second mock-up shown in Fig. 1(b) was intended to represent the lighting scenario in a contemporary retail clothing store. Two female and two male mannequins were dressed in a certain variety of decent styles, colors, and patterns to imitate a typical display area.
Fig. 1. The two mock-ups used for testing the models. (a) Mock-up 1 represents a typical fruit and vegetable supermarket counter. (b) Mock-up 2 represents a contemporary retail clothing store.
The two experiments took place on separate days, distributed over the first and second half of a single experimental week. In both cases, natural daylight was blocked from entry. The room illumination was provided by a system of six tunable-white Lunexo LED luminaires (TRILUX GmbH & Co. KG, Arnsberg, Germany) that were suspended from the ceiling at a height of 2.6 m. Their corresponding layout in relation to the room dimensions is depicted in Fig. 2. The luminaires can natively be operated in the range from 2700 K to 6500 K and offer a controllable direct/indirect light distribution ratio. Their indirect component was further extended by custom-made light engines comprising five monochromatic LED channels (blue 454 nm, cyan 476 nm, green 540 nm, amber 595 nm, and red 648 nm) that were additionally attached to each luminaire and connected to the joint DALI bus for an overall lighting control. Both the direct and the upgraded indirect components of the luminaires offered Lambertian emission characteristics so that a homogeneous illumination of the room and, in particular, of the displayed objects could be achieved.
Fig. 2. Room layout and luminaire distribution. Additionally shown are the approximate standing positions of the subjects during the experimental rating sessions. A maximum of four subjects were tested at the same time.
For the experiments, the direct/indirect ratio was set to a fixed value of $\alpha = 0.7$, which according to Houser et al. [37] represents a user-preferred value for downlight/uplight combinations. Using the model of Khanh et al., thirteen different light spectra were optimized in such a way that an approximately equidistant distribution of metric values was achieved on the respective sub-scales of brightness B, visual clarity VC, and color preference CP. With regard to lighting practice, this basically represents the use of the provided mathematical formalism to generate spectra of predefined model prediction values, which is comparable to optimizing light spectra for reaching a specific color rendering index (CRI). The corresponding SPDs are visualized in Fig. 3. In addition, Table 1 summarizes the (most relevant) photometric and (some selected) colorimetric properties of these thirteen test conditions as well as their B, VC, and CP metric values. Note that scene preference SP was not considered in this work since this attribute usually refers to the entire lit interior with a much stronger focus on room perception than it could be covered by the two mock-up settings. Another validation experiment specifically designed for this model aspect is therefore on its way and results should be reported soon.
Table 1. Model predictions for B, VC and CP as well as a selection of photometric quantities and color quality metrics of the thirteen SPDs used in both mock-up scenarios
2.3 Participants and test procedure
In total, 59 volunteers were recruited through email notices, electronic and conventional postings, and word-of-mouth. Potential candidates were excluded from the study only if they suffered from a known color vision deficiency or any other severe vision impairment. In particular, wearing glasses or contact lenses for visual acuity correction was no reason for exclusion as long as the limits of $\pm {6}\,\textrm {dpt}$ or $\pm {4}\,\textrm {dpt}$ for astigmatism were not exceeded. All participants were paid and written consent was obtained in each case. The study was approved by the ethical review committee of the Technical University of Darmstadt. All experiments, data collection and data storage were conducted in accordance with national and international ethical standards and, in particular, adhere to the Declaration of Helsinki and the requirements of the German Research Foundation (DFG).
The participants were randomly assigned to either of the two different mock-up settings so that 30 subjects (12 female, 19 male) between 20 and 59 years ($\emptyset = 27.2\pm 10.8$) rated the test lighting conditions in the supermarket scenario, while 29 subjects (14 female, 15 male) between 15 and 51 years ($\emptyset = 26.3\pm 7.0$) assessed the same lighting conditions in the retail context. The two different test sessions took place on separate days during the first and second half of a single experimental week. Subjects within each session were assigned to different experimental time slots per day depending on their personal schedule and preferences. A maximum of four participants were tested at the same time. Fig. 2 depicts their approximate rating positions in relation to the displayed test objects. A random assignment of the participants to the individual rating positions was performed.
For each mock-up setting, the test protocol comprised three consecutive evaluation parts, one for each of the three different perceptual attributes considered in this work. The order in which subjects were asked to give their ratings with respect to these attributes was fixed and started with the assessment of color preference CP followed by visual clarity VC and, finally, perceived brightness B. In total, it took the participants 1.5 h to complete the experimental session including short breaks between the three individual parts.
Ratings for the different perceptual attributes were collected by applying the same rating instruments as used by Khanh et al. [43,44] in their original study for the development of the respective models. These were a non-uniform Brückner scale (see Fig. 4, left) for the assessments of CP and VC as well as a continuous linear percentage scale with an equidistant partitioning for collecting judgements of perceived brightness (see Fig. 4, right). The non-uniform Brückner scale with semantic categories was specifically designed as an intuitive tool for the evaluation of color quality and color-related aspects in lighting that is particularly suitable for inexperienced observers as encountered in the present case. For further details on its derivation, the interested reader is referred to Bodrogi et al. [82,83]. Linear or equidistant categorical rating scales on the other hand are the established method of choice for the assessment of perceived brightness [84]. Both types of scales had been applied successfully in previous research by some of the authors [25–29] and proven valid instruments for the proper evaluation of the observers’ subjective impressions of object appearance.
Fig. 4. Rating scales used for data collection. (left) Non-uniform Brückner rating scale for the assessment of color preference and visual clarity, labelled by the categories ’excellent’ (97.9), ’very good’ (91.6), ’good’ (79.6), ’moderate’ (52.9), ’poor’ (41.2), ’bad’ (26.5), and ’very bad’ (12.8). (right) Percentage scale for collecting the ratings of perceived brightness with an equidistant partitioning.
Starting with the assessment of color preference CP, subjects were first explained the concept behind this attribute. Furthermore, dedicated anchor stimuli of low (SPD 1) and high (SPD 11) CP predictions were presented to them in order to get an idea of the visual range in which corresponding changes of object appearance could be expected. Note that avoiding to show the stimulus with the highest prediction value (i.e., SPD 13) occurred intentionally. This decision was made to not bias the subjects’ ratings too much towards the upper end of the respective rating scale but, at the same time, still provide a comprehensible visualization of the different model attributes and their expected effects on the perceived objects’ color appearance.
The Subjects were then instructed to rate their degree of visual appreciation (or disapproval) of the color appearance of the mock-up objects for each of the thirteen tailored test light conditions. The order in which these lighting conditions were presented for rating was completely randomized. Changing the lighting was always announced by a computer-generated voice asking the participants to close their eyes before the transition process from the current to the next SPD was initiated. Compliance was controlled by the experimenter. The end of the transition between two light spectra was indicated by a sound signal, followed by a 100 s adaptation period in which the participants were instructed to embrace the whole scene and get an overall impression of the objects’ appearance before thinking of a specific rating. After the adaptation time had passed, the computer-generated voice announced the beginning of the rating period. Subjects were then given another 20 s to indicate their CP rating for the current lighting situation on the corresponding rating scale before switching to the next lighting condition.
This test protocol was essentially repeated for the remaining two attributes, with the sole difference that during the training for perceived brightness two additional anchors of no light and maximum light (SPD 13) were presented to the subjects. For assessing visual clarity VC, subjects were instructed to indicate how well or poorly the presented test objects are rendered by the different lighting conditions with regard to their color fine structure, contrasts, and shadings, whereas for the assessment of perceived brightness B, they were asked to rate how bright or dark the respective mock-up scenario appears under each of the different light settings in relation to the minimum and maximum anchor stimuli presented during the training.
2.4 Performance and statistical analysis
The applicability of Khanh et al.’s lighting quality model was analyzed in terms of predictive performance by calculating the correlations between the model predictions for each perceptual attribute and the corresponding subject ratings obtain for the two different mock-up scenarios. As can be seen from Eqs. (1) to (5), the illuminance $E_{\mathrm {v}}$ provides the main contribution to all three attributes of Khanh et al.’s model. This basically reflects the fact that higher light levels in general lead to a more vivid representation of object colors, an increase in perceived brightness, and a better distinction of the objects’ color fine structure and, thus, result in higher ratings of the respective appearance attributes. It is therefore expected that the subjects’ judgements are mainly affected by the level of light intensity with a high correlation between the different model attributes.
Comparisons of model performance were conducted with regard to a selection of other lighting quality metrics known from literature. For CP, these included the CIE Color Rendering Index (CRI: $R_{\mathrm {a}}$) [85], its updated version $\mathrm {CRI}_{\mathrm {2012}}$ [86], the Color Quality Scale (CQS: General Scale $Q_{\mathrm {a}}$, Color Fidelity Scale $Q_{\mathrm {f}}$, Gamut Area Scale $Q_{\mathrm {g}}$ and Color Preference Scale $Q_{\mathrm {p}}$) [73], the Feeling of Contrast Index (FCI) [87], the Memory Color Rendering Index (MCRI: $R_{\mathrm {m}}$) [88], the Memory Color Preference Index (MCPI) [81], and the IES Color Fidelity Score $R_{\mathrm {f}}$ and Color Gamut Score $R_{\mathrm {g}}$ [89]. For the brightness attribute B, corresponding model prediction correlations are compared to those obtained for the Fotios-Levermore [62] and the Sagawa [63] brightness models. For the evaluation of VC, all the aforementioned metrics and models were included. In addition, the subjects’ ratings for each perceptual attribute were correlated with the measured illuminance $E_{\mathrm {v}}$.
Statistical analysis was performed in R using linear mixed-effects models, where participant was considered as a random factor, while the light setting, the type of mock-up, the gender, and the age of the subjects were treated as fixed factors. In all cases, statistical significance was concluded for reported p-values smaller than $0.05$.
3. Results
The subjects’ mean ratings of B, VC, and CP for each lighting condition averaged over both mock-up scenarios are summarized in Table 2 together with their respective standard deviations (SDs). In addition, Fig. 5 relates the subjects’ mean ratings obtained for the different perceptual attributes to the corresponding model prediction values.
Fig. 5. Subjects' mean ratings for (a) perceived brightness, (b) visual clarity, and (c) color preference in comparison to their respective model predictions. Error bars indicate corresponding SDs. In addition, linear regression results (solid red lines) are compared to the expectations in case of ideal correlations (dashed purplish lines).
Table 2. Subjects’ mean ratings and standard deviations for the perceptual attributes of perceived brightness (B), visual clarity (VC), and color preference (CP) of the 13 test light conditions averaged over both mock-up scenarios
As can be seen, model predictions for all three perceptual attributes show an excellent linear correlation with the subjects’ mean ratings averaged over both mock-up scenarios. As expected, a high correlation with $E_{\mathrm {v}}$ could be confirmed. The corresponding Pearson $r_{\mathrm {p}}$ and Spearman $r_{\mathrm {s}}$ correlation coefficients are summarized in Table 3 and compared to those obtained for the various lighting quality metrics discussed in Sec. 2.4. In all cases, Khanh et al.’s model shows the largest overall correlations and, in particular with regard to VC and CP, clearly outperforms the alternatives considered in this work, most of which even fail to predict the light sources’ rank order correctly. To be fair, it should be noted that, in contrast to VC and CP, none of these metrics have been designed to evaluate and compare light sources at different illuminance levels $E_{\mathrm {v}}$, so that their poor correlation performance is not really surprising. Nonetheless, some of these metrics are still widely used in practice to compare the color rendition properties of light sources regardless of their level of illuminance. The present comparison therefore addresses the lighting practitioner by clearly showing the necessity of an illuminance dependent modelling for an adequate description of lighting quality.
Table 3. Pearson $r_{\mathrm {p}}$ and Spearman $r_{\mathrm {s}}$ correlation coefficients describing the relation between the various model and metric predictions and the subjects’ mean ratings for perceived brightness B, visual clarity VC, and color preference CP
The results of the statistical analysis performed separately for each perceptual attribute are shown in Table 4. Bold entries indicate a statistically significant effect of the respective predictor or interaction term specified in the first column. Regarding the subjects’ brightness ratings, there are significant main effects of lighting condition, $\chi ^2(12) = 1024.1$, $p<0.001$, and age, $\chi ^2(1) = 4.7$, $p=0.029$, and a significant interaction between lighting condition and age, $\chi ^2(12) = 31.1$, $p=0.002$, indicating that ratings for different lighting conditions differ between participants of different ages. No significant differences, on the other hand, are found between the two mock-up scenarios, nor are there any further significant higher-order interaction terms. For the ratings of visual clarity, significant main effects of lighting condition, $\chi ^2(12) = 991.9$, $p<0.001$, and mock-up, $\chi ^2(1) = 4.6$, $p=0.032$, can be stated. In addition, there is a statistically significant interaction effect of lighting condition and mock-up, $\chi ^2(12) = 30.4$, $p=0.003$, suggesting that subject ratings for different lighting conditions differ between both mock-up scenarios. Other than that, there are no further main or interaction effects to be reported. Finally, for the ratings of color preference, there are again significant main effects of lighting condition, $\chi ^2(12) = 381.9$, $p<0.001$, and mock-up, $\chi ^2(1) = 4.1$, $p=0.044$. Moreover, a significant interaction effect between lighting condition and gender has been revealed, $\chi ^2(12) = 27.3$, $p=0.007$. No further main effects or higher-order interaction terms are found to be significant.
Table 4. Linear mixed-effects model results for the B, VC, and CP outcome measures. Bold numbers indicated a statistically significant effect of the respective predictor or interaction term specified in the first column.
4. Discussion
From the previous section, it is evident that Khanh et al.’s model features an excellent predictive performance. Almost perfect correlations between model predictions and subjective ratings are observed for the relevant perceptual attributes of perceived brightness, visual clarity, and color preference. Except for the brightness models of Sagawa and Fotios-Levermore, which also showed good to excellent correlations for visual clarity and perceived brightness, none of the remaining color quality metrics considered in this work were capable of adequately modeling the subjects’ ratings. For instance, even the preference-related color quality metrics such as $Q_{\mathrm {p}}$, Smet et al.’s $R_{\mathrm {m}}$, or Babilon et al.’s MCPI failed to predict color preference. This can be explained by the fact that all these metrics lack the inclusion of an illuminance component and therefore are not intended to be used for the evaluation of light sources of different $E_{\mathrm {v}}$. However, as shown by Wei et al. [46,90], changes of light levels at otherwise identical lighting conditions considerably affect the objects’ color appearance, which, as discussed in the introduction section, relates to the Hunt effect [47]. Changes in the perceived color appearance in turn lead to variations in the subjects’ preference ratings when light levels are increased or decreased and, therefore, require an illuminance dependent modelling for an adequate and comprehensive description of lighting quality, as it was performed by Khanh and his co-workers.
Despite the resulting excellent correlations between their model predictions and the subjects’ mean ratings, some slight deviations from an ideal model can still be identified. Theoretically, an ideal model should predict metric values for the individual perceptual attributes that exactly tally with the experimental observations, i.e., when the subjects’ mean rating for a given lighting condition is for example 60 on any of the perceptual sub-scales, the respective model should have also prognosticated a value of 60. However, as can be seen from Fig. 5, there are some marked differences between the regression results fit to the subjects’ attribute-specific rating data (solid red lines) and the correspondingly assumed ideal model predictions (dashed purplish lines). Comparing subject ratings and model predictions at each level of any of the perceptual attributes considered in this work, a model overestimation for high values and a respective underestimation for low values can be noticed. These findings can be explained by the occurrence of two complementary effects. The first one is a non-negligible central tendency bias. Subjects in general tend to place most of their ratings in the middle range of a rating scale that is anchored in its upper and lower values by some categorical (semantic) or numerical extremes, like in the present case. This basically yields a compression in the subjects’ ratings leading to mean values for large (small) model predictions that are slightly smaller (larger) than expected by the model. Note that this was not an issue in Khanh et al.’s original work as they optimized the model parameters to match the model predictions with the respective subject ratings, ignoring any potential rating-scale bias. The second aspect that plays a role is the fact that a light spectrum that would have scored a certain value on any of the perceptual sub-scales in their original experiments does not necessarily induce the same rating when being assessed in a new environmental context. Thus, instead of being accessible in an absolute manner, perceptual attributes can only be judged in relation to the current viewing conditions and, therefore, might lead to distinct ratings for different situations.
However, from a practical point of view, where stakeholders often have to decide between several given options to select the most appropriate lighting solution for a specific application context, an ideal model performance is not really necessary. Usually, it is sufficient to know the light sources’ rank order with regard to a selection of relevant features to gain an idea which of the potential lighting options scores best and identifies as the most adequate solution. Under this aspect, Khanh et al.’s model provides an outstandingly excellent performance as it is not only capable of correctly predicting the light sources’ perceptual rank order in various different lighting contexts, see also Refs. [71,72], it even shows perfectly linear correlations with the subjects’ respective mean ratings. Thus, supplementing the preliminary evidence from literature, the reported results clearly emphasize the suitability and validity of Khanh et al.’s multi-dimensional approach and model formalism to be applied for the planning and optimization of modern lighting solutions intended to address the users’ perceptual needs.
Besides providing evidence for the practical relevance and validity of Khanh et al.’s model predictions, the results of the present study further implied some interesting statistical interferences. Firstly, a significant main effect of mock-up was found for the perceptual attributes of visual clarity and color preference, but not for perceived brightness. This indicates that the subjects’ general perception of brightness is less affected by the nature and chromatic characteristics of the viewing content than their perception of visual clarity or color preference. This finding is not surprising given the fact that perceived brightness mainly assesses an achromatic light-dark impression, whereas judgements of visual clarity and color preference explicitly ask to include color information. Secondly, the statistical analysis revealed some significant interaction effects. The break-down of these effects is illustrated in Fig. 6.
Fig. 6. Break-down of significant interaction effects. (a) Mean subject ratings vs. model predictions for perceived brightness as observed for younger (<30 years) compared to older ($\geq$30 years) subjects. This division into two equally-sized subgroups was performed for visualization purposes only. (b) Mean subject ratings vs. model predictions for visual clarity as obtained for both mock-up scenarios. (c) Mean ratings of females and males vs. model predictions for color preference. In all cases, error bars indicate corresponding SDs.
Regarding the ratings of perceived brightness, there was a significant lighting condition $\times$ age interaction. For the visualization of this effect, the sample was divided into two approximately equally-sized subgroups of older ($\geq$30 years) versus younger (<30 years) participants. As can be seen from Fig. 6(a), older participants on average rated dimmer light settings brighter than their younger counterparts, whereas no differences between the corresponding mean ratings were observed for the brighter light settings. Re-performing the statistical analysis using age group (i.e., older and younger than 30 years) as a predictor in the model leads to a significant lighting condition $\times$ age group interaction effect for brightness perception, $\chi ^2(12) = 22.4$, $p=0.033$.
These findings might be explained by age-related changes of the visual system. During training (see Sec. 2.3), the two anchor stimuli of no light and maximum light, accompanied by a corresponding hint for the latter, were presented to both age groups. This defined a clear reference especially for the upper end of the applied rating scale and, thus, yielded similar results for the brighter light settings. However, due to age-related changes in the transmission properties of the human eye lens showing a significant reduction in the spectral region of shorter wavelengths [91–94], it can be expected that, despite their comparable brightness ratings, the total amount of light received by the retina is considerably decreased in older compared to younger participants for light settings at the upper end of the respective rating scale. For dimmer light settings, on the other hand, there is not only a reduction of light levels but, following Khanh et al.’s brightness model, also a shift to warmer light spectra, reducing the contributions of short wavelengths. Because of the relative nature of brightness ratings, this in turn increases the perceived brightness of dimmer light settings for older subjects relative to the upper anchor stimulus and eventually results in the observed larger ratings for these lighting conditions compared to those obtained for the younger subjects. Note that despite providing a possible explanation for the observed results, it still remains unclear from the current data whether lens aging in the elderly is the relevant process here and responsible for the rating differences between older and younger subjects. More systematic experiments including the explicit determination of the individual observer’s lens aging state are required to draw final conclusions.
For visual clarity, Fig. 6(b) breaks down the corresponding lighting condition $\times$ mock-up effect. As can be seen, the supermarket setting (1st mock-up) compared to the retail clothing store setting (2nd mock-up) yields slightly larger mean ratings for low model predictions, while for large model predictions equal subject ratings are observed. Comparing both mock-up settings, chromatic and achromatic contrasts are more pronounced and clearly more apparent in the second case due to the patterns of the displayed clothes and their respective fabric structures, whereas the different vegetables and fruits of the supermarket scenario showed rather smooth surfaces of subtle color gradients. These differences might be an explanation for the observed results as especially for the dimmer light settings some subjects reported difficulties in discerning different levels of visual clarity between the different lighting conditions for the supermarket scenario but not for the retail clothing store setting, where changes in visual clarity seemed to be better noticeable even for the dimmer light settings.
Finally, the results for color preference showed a significant lighting condition $\times$ gender interaction, which is illustrated in Fig. 6(c). A considerably stronger compression of the ratings is observed for female compared to male participants. From the data, it is not a hundred percent clear whether this compression is caused by a stronger central tendency bias in female subjects or if there is a general difference in their perception of color preference that, in comparison to the male subjects, leads to a stronger model over(under)estimation for high (low) CP values. What militates against the latter is that there was no significant main effect of gender. This indicates that when neglecting all other effects there was no overall difference in the color preference ratings between male and female subjects. In addition, having the intersection of both regression lines of the two sub-groups appear approximately in the middle of the corresponding rating scale further hints at an enhanced central tendency bias for female raters rather than at gender-dependent variations in the assessments of color preference.
In summary, it should be noted that despite the reported interactions that may cause significant differences in the ratings of the various light settings depending on the level of another variable, the practically relevant rank order correlations are hardly affected by these interactions. In all cases shown in Fig. 6, corresponding Pearson and Spearman correlation coefficients are always found to be larger than 0.96. These excellent correlation results again emphasize the model’s capability of correctly predicting the light sources’ perceptual rank order, independent of whether or not subject ratings scale differently because of some interference from non-negligible interaction effects.
5. Conclusions
Adjusting lighting conditions in such a way that they should match the users’ preferences requires appropriate tools for predicting how changes of the lit environment affect human perception. In this context, Khanh et al. [43,44] recently proposed a new model formalism for the lighting practitioner, intended to be applied for the planning and optimization of modern lighting solutions while taking into account the users’ perceptual needs. Building up on their proposal, the present work provided a proof of the model’s applicability from a lighting practitioners point of view with regard to some realistic use-case scenarios. For this purpose, two dedicated experiments representing different lighting contexts were conducted and their results were discussed accordingly. From the analysis of the study data, excellent correlations between model predictions and the subjects’ mean ratings for the different model sub-scales of perceived brightness, visual clarity, and color preference could be observed. From a practical point of view, the model’s capability of correctly predicting the light sources’ perceptual rank order across a large variety of different modalities appears to be its most important feature and, thus, clearly emphasizes its relevance for application.
In addition to these model-specific conclusions, the present work also underlines the importance of pursuing a multi-dimensional approach for a comprehensive description of lighting quality. In particular, it was again affirmed that the inclusion of a light-level dependent component is an essential necessity for properly modelling perceptual attributes in lighting. Moreover, interaction effects between illuminance, CCT, and object saturation need to be considered explicitly as they considerably affect how people perceive the lit environment. Thus, the new model formalism proposed by Khanh et al. provides an excellent first step in the right direction towards a more user-oriented design of modern lighting solutions.
Funding
Technische Universität Darmstadt (Forum for Interdisciplinary Research 2019-4, Ernst Ludwig Mobility Grant); Deutsche Forschungsgemeinschaft (445336968).
Author Contributions. Conceptualization, S.Ba., J.K., S.Be., and B.J.; methodology, S.Ba. and J.K.; software, J.K. and Q.V.T.; spectral optimization, Q.V.T; running the studies, J.K., S.Be, and S.Br.; formal analysis, S.Ba. and J.K.; data curation, J.K. and S.Ba; writing–original draft preparation, S.Ba. and J.K.; writing–review and editing, S.Ba., J.K., S.Br., B.J., A.M.J., Q.V.T., and T.Q.K.; visualization, J.K. and Q.V.T.; supervision, T.Q.K. and A.M.J. All authors have read and agreed to the published version of the manuscript.
Disclosures
The authors declare no conflict of interest.
Data availability
All data generated or analyzed to support the findings of the present study are included in this article. The raw data can be obtained from the authors, upon reasonable request.
References
1. P. Boyce, “Editorial: Exploring human-centric lighting,” Light. Res. Technol. 48(2), 101 (2016). [CrossRef]
2. K. W. Houser, P. R. Boyce, J. M. Zeitzer, and M. Herf, “Human-centric lighting: Myth, magic or metaphor?” Light. Res. Technol. 53(2), 97–118 (2021). [CrossRef]
3. S. Babilon, S. Beck, J. Kunkel, J. Klabes, P. Myland, S. Benkner, and T. Q. Khanh, “Measurement of circadian effectiveness in lighting for office applications,” Appl. Sci. 11(15), 6936 (2021). [CrossRef]
4. S. Babilon, J. Lenz, S. Beck, P. Myland, J. Klabes, S. Klir, and T. Q. Khanh, “Task-related luminance distributions for office lighting scenarios,” Light. Eng. 29, 115–128 (2021). [CrossRef]
5. P. Khademagha, M. B. C. Aries, A. L. P. Rosemann, and E. J. van Loenen, “Implementing non-image-forming effects of light in the built environment: A review on what we need,” Build. Environ. 108, 263–272 (2016). [CrossRef]
6. S. Westland, Q. Pan, and S. Lee, “A review of the effects of colour and light on non-image function in humans,” Color. Technol. 133, 349–361 (2017). [CrossRef]
7. R. Lok, K. C. H. J. Smolders, D. G. M. Beersma, and Y. A. W. de Kort, “Light, alertness, and alerting effects of white light: A literature overview,” J. Biol. Rhythm. 33(6), 589–601 (2018). [CrossRef]
8. J. L. Souman, A. M. Tinga, S. F. te Pas, R. van Ee, and B. N. S. Vlaskamp, “Acute alerting effects of light: A systematic literature review,” Behav. Brain Res. 337, 228–239 (2018). [CrossRef]
9. S. Fotios and D. Atli, “Comparing judgements of visual clarity and spatial brightness through an analysis of studies using the category rating procedure,” Leukos 8(4), 261–281 (2012). [CrossRef]
10. M. S. Rea, X. Mou, and J. D. Bullough, “Scene brightness of illuminated interiors,” Light. Res. Technol. 48(7), 823–831 (2016). [CrossRef]
11. J. Duff, K. Kelly, and C. Cuttle, “Perceived adequacy of illumination, spatial brightness, horizontal illuminance and mean room surface exitance in a small office,” Light. Res. Technol. 49(2), 133–146 (2017). [CrossRef]
12. S. M. Aston and H. E. Belichambers, “Illumination, colour rendering and visual clarity,” Light. Res. Technol. 1(4), 259–261 (1969). [CrossRef]
13. H. E. Belichambers and A. C. Godby, “Illumination, colour rendering and visual clarity,” Light. Res. Technol. 4(2), 104–106 (1972). [CrossRef]
14. W. A. Thornton and E. Chen, “What is visual clarity?” J. Illum. Eng. Soc. 7(2), 85–94 (1978). [CrossRef]
15. P. R. Boyce and C. Cuttle, “Effect of correlated colour temperature on the perception of interiors and colour discrimination performance,” Light. Res. Technol. 22(1), 19–36 (1990). [CrossRef]
16. K. Hashimoto and Y. Nayatani, “Visual clarity and feeling of contrast,” Color Res. Appl. 19, 171–185 (1994). [CrossRef]
17. P. L. Vrabel, C. A. Bernecker, and R. G. Mistrick, “Visual performance and visual clarity under electric light sources: Part II–Visual clarity,” J. Illum. Eng. Soc. 27(1), 29–41 (1998). [CrossRef]
18. M. S. Rea and J. P. Freyssinier, “White lighting,” Color Research and Application 38, 82–92 (2013). [CrossRef]
19. P. Bodrogi, X. Guo, D. Stojanovic, S. Fischer, and T. Q. Khanh, “Observer preference for perceived illumination chromaticity,” Color Res. Appl. 43, 506–516 (2018). [CrossRef]
20. Y. Ohno and M. Fein, “Vision experiment on acceptable and preferred white light chromaticity for lighting,” in Proceedings of CIE 2014 Lighting Quality and Energy Efficiency, (CIE, Kuala Lumpur, Malaysia, 2014), pp. 192–199.
21. M. Perz, R. Baselmans, and D. Sekulovski, “Perception of illumination whiteness,” in Proceedings of the 4th CIE Expert Symposium on Colour and Visual Appearance, (CIE, Prague, Czech Republic, 2016), pp. 1–7.
22. Y. Lin, X. Xiao, P. Bodrogi, and T. Q. Khanh, “Intercultural colour temperature preference (2300 k - 5800 k) of chinese and european subjects for white objects,” in Proceedings of the 28th CIE Session, (CIE, Manchester, UK, 2015), pp. 928–932.
23. P. Bodrogi, T. Q. Khanh, D. Stojanovic, and Y. Lin, “Intercultural colour temperature preference of chinese and european subjects living in germany,” Light. Eng. 24, 8–11 (2016).
24. P. Bodrogi, Y. Lin, X. Xiao, D. Stojanovic, and T. Q. Khanh, “Intercultural observer preference for perceived illumination chromaticity for different coloured object scenes,” Light. Res. Technol. 49(3), 305–315 (2017). [CrossRef]
25. T. Q. Khanh, P. Bodrogi, Q. T. Vinh, and D. Stojanovic, “Colour preference, naturalness, vividness and colour quality metrics, Part 1: Experiments in a room,” Light. Res. Technol. 49(6), 697–713 (2017). [CrossRef]
26. T. Q. Khanh, P. Bodrogi, Q. T. Vinh, and D. Stojanovic, “Colour preference, naturalness, vividness and colour quality metrics, Part 2: Experiments in a viewing booth and analysis of the combined dataset,” Light. Res. Technol. 49(6), 714–726 (2017). [CrossRef]
27. T. Q. Khanh and P. Bodrogi, “Colour preference, naturalness, vividness and colour quality metrics, Part 3: Experiments with makeup products and analysis of the complete warm white dataset,” Light. Res. Technol. 50(2), 218–236 (2018). [CrossRef]
28. T. Q. Khanh, P. Bodrogi, Q. T. Vinh, X. Guo, and T. T. Anh, “Colour preference, naturalness, vividness and colour quality metrics, Part 4: Experiments with still life arrangements at different correlated colour temperatures,” Light. Res. Technol. 50(6), 862–879 (2018). [CrossRef]
29. T. Q. Khanh, P. Bodrogi, X. Guo, Q. T. Vinh, and S. Fischer, “Colour preference, naturalness, vividness and colour quality metrics, Part 5: A colour preference experiment at 2000 lx in a real room,” Light. Res. Technol. 51(2), 262–279 (2019). [CrossRef]
30. J. E. Flynn, T. J. Spencer, O. Martyniuk, and C. Hendrick, “Interim study of procedures for investigating the effect of light on impression and behavior,” J. Illum. Eng. Soc. 3(1), 87–94 (1973). [CrossRef]
31. P. R. Tregenza, S. M. Romaya, S. P. Dawe, L. J. Heap, and B. Tuck, “Consistency and variation in preferences for office lighting,” Light. Res. Technol. 6(4), 205–211 (1974). [CrossRef]
32. K. F. Roll and H. J. Hentschel, “Luminance patterns in interiors and balanced perception,” in Proceedings of the 21th CIE Session, (CIE, Venice, Italy, 1987), pp. 194–196.
33. H. J. Hentschel, E. Klein, J. Leibig, and K. F. Roll, “Energy-effective direct/indirect office and VDU-lighting systems: Test and application,” J. Illum. Eng. Soc. 16(2), 89–105 (1987). [CrossRef]
34. D. L. Loe, K. P. Mansfield, and E. Rowlands, “Appearance of lit environment and its relevance in lighting design: Experimental study,” Light. Res. Technol. 26(3), 119–133 (1994). [CrossRef]
35. D. K. Tiller and J. A. Veitch, “Perceived room brightness: Pilot study on the effect of luminance distribution,” Light. Res. Technol. 27(2), 93–101 (1995). [CrossRef]
36. S. Fleischer, H. Krueger, and C. Schierz, “Effect of brightness distribution and light colours on office staff: Results of the Lighting Harmony project,” in Proceedings of 9th European Lighting Conference Lux Europe 2001, (Reykjavík, Iceland, 2001), pp. 76–80.
37. K. W. Houser, D. K. Tiller, C. A. Bernecker, and R. G. Mistrick, “The subjective response to linear fluorescent direct/indirect lighting systems,” Light. Res. Technol. 34(3), 243–260 (2002). [CrossRef]
38. K. I. Fostervold and J. Nersveen, “Proportions of direct and indirect indoor lighting – The effect on health, well-being and cognitive performance of office workers,” Light. Res. Technol. 40(3), 175–200 (2008). [CrossRef]
39. H. J. A. de Vries, I. E. J. Heynderickx, Y. A. W. de Kort, and B. de Ruyter, “Wall illumination – Beyond room appraisal,” in Proceedings of the 28th CIE Session, (CIE, Manchester, UK, 2015), pp. 284–290.
40. J. T. Sullivan and M. Donn, “Light distribution and spatial brightness: Relative importance of the walls, ceiling, and floor,” in Proceedings of the CIE 2016 Lighting Quality and Energy Efficiency Conference, (CIE, Melbourne, Australia, 2016), pp. 59–69.
41. S. Chraibi, L. Crommentuijn, E. van Loenen, and A. Rosemann, “Influence of wall luminance and uniformity on preferred task illuminance,” Build. Environ. 117, 24–35 (2017). [CrossRef]
42. C. de Bakker, M. Aarts, H. Kort, E. van Loenen, and A. Rosemann, “Preferred luminance distributions in open-plan offices in relation to time-of-day and subjective alertness,” Leukos 17(1), 3–20 (2021). [CrossRef]
43. T. Q. Khanh, P. Bodrogi, X. Guo, and P. Q. Anh, “Towards a user preference model for interior lighting Part 1: Concept of the user preference model and experimental method,” Light. Res. Technol. 51(7), 1014–1029 (2019). [CrossRef]
44. T. Q. Khanh, P. Bodrogi, X. Guo, and P. Q. Anh, “Towards a user preference model for interior lighting. Part 2: Experimental results and modelling,” Light. Res. Technol. 51(7), 1030–1043 (2019). [CrossRef]
45. Y. Kawashima and Y. Ohno, “Vision experiment on verification of hunt effect in lighting,” in Proceedings of the 29th Quadrennial Session of the CIE, (International Commission on Illumination, CIE, 2019).
46. M. Wei, W. Bao, and H. Huang, “Consideration of light level in specifying light source color rendition,” Leukos 16(1), 55–65 (2020). [CrossRef]
47. R. W. G. Hunt, “Light and Dark Adaptation and Perception of Color,” J. Opt. Soc. Am. 42(3), 190–199 (1952). [CrossRef]
48. J. J. Balder, “Erwünschte Leuchtdichten in Büroräumen (Preferred Luminance in Offices),” Lichttechnik 9, 455–461 (1957).
49. E. Muck and H. W. Bodmann, “Die Bedeutung des Beleuchtungsniveaus bei praktischer Sehtätigkeit,” Lichttechnik 13, 502–507 (1961).
50. J. M. Westhoff and H. W. Horemann, “Contribution to Bodmann,” Lichttechnik 14, 24 (1963).
51. H. W. Bodmann, G. Söllner, and E. Voit, “Bewertung von Beleuchtungsanlagen bei verschiedenen Lichtarten,” in Proceedings of the CIE 1963 Conference, (International Commission on Illumination CIE, Vienna, Austria, 1963), pp. 502–509.
52. H. W. Bodmann, “Quality of interior lighting based on luminance,” Light. Res. Technol. 32(1 IEStrans), 22–40 (1967). [CrossRef]
53. W. Riemenschneider, “Beleuchtungsstärken für Arbeiträume,” Bull. des Verbands Schweizerischer Elektrizitätsunternehmen 58, 19–25 (1967).
54. P. R. Boyce, “The influence of illumination level on prolonged work performance,” Light. Res. Technol. 2(2), 74–94 (1970). [CrossRef]
55. D. Fischer, “Optimale Beleuchtungsniveaus in Arbeitsräumen Teil 1,” Lichttechnik 22, 61–63 (1970).
56. D. Fischer, “Optimale Beleuchtungsniveaus in Arbeitsräumen Teil 2,” Lichttechnik 22, 103–105 (1970).
57. S. H. A. Begemann, G. J. van den Beld, and A. D. Tenner, “Daylight, artificial light and people in an office environment, overview of visual and biological responses,” Int. J. Ind. Ergonomics 20(3), 231–239 (1997). [CrossRef]
58. C. Moosmann, Visueller Komfort und Tageslicht am Büroarbeitsplatz. Eine Felduntersuchung in neun Gebäuden (KIT Scientific Publishing, 2014).
59. R. M. Boynton, “A system of photometry and colorimetry based on cone excitations,” Color Res. Appl. 11, 244–252 (1986). [CrossRef]
60. J. Ju, D. Chen, and Y. Lin, “Effects of correlated color temperature on spatial brightness perception,” Color Res. Appl. 37, 450–454 (2012). [CrossRef]
61. L. E. Marks, “Blue sensitive cones can mediate brightness,” Vis. Res. 14(12), 1493–1494 (1974). [CrossRef]
62. S. Fotios and G. Levermore, “Chromatic effect on apparent brightness in interior spaces II: sws lumens model,” Light. Res. Technol. 30(3), 103–106 (1998). [CrossRef]
63. K. Sagawa, “Toward a CIE supplementary system of photometry: Brightness at any level including mesopic vision,” Oph. Phys. Opt. 26, 240–245 (2006). [CrossRef]
64. M. Yamakawa, S. I. Tsujimura, and K. Okajima, “A quantitative analysis of the contribution of melanopsin to brightness perception,” Sci. Rep. 9(1), 7568 (2019). [CrossRef]
65. P. Bodrogi, X. Guo, and T. Q. Khanh, “Brightness in the photopic range: Psychophysical modelling with blue-sensitive retinal signals,” Light. Eng. 28, 9–24 (2020). [CrossRef]
66. Z. Huang, Q. Liu, S. Westland, M. Pointer, M. R. Luo, and K. Xiao, “Light dominates colour preference when correlated colour temperature differs,” Light. Res. Technol. 50(7), 995–1012 (2018). [CrossRef]
67. Z. Huang, Q. Liu, Y. Liu, M. Pointer, M. Luo, Q. Wang, and B. Wu, “Best lighting for jeans, part 1: Optimising colour preference and colour discrimination with multiple correlated colour temperatures,” Light. Res. Technol. 51(8), 1208–1223 (2019). [CrossRef]
68. Q. T. Vinh, P. Bodrogi, T. Q. Khanh, and T. T. Anh, “Colour preference depends on colour temperature, illuminance level and object saturation – A new metric,” Light. Eng. 27, 137–151 (2019). [CrossRef]
69. T. Q. Khanh, P. Bodrogi, and T. Q. Vinh, “Visual colour quality experiments at the Technische Universität Darmstadt,” in Colour Quality of Semiconductor and Conventional Light Sources, (Wiley-VCH, Weinheim, Germany, 2016.
70. P. Bodrogi and T. Q. Khanh, “Visual clarity and brightness in indoor and outdoor lighting: Experiments and modelling,” in Proceedings of the CIE Midterm Meeting 2017, (CIE, Jeju, Republic of Korea, 2017).
71. Z. Wang, Y. Nagai, J. Liu, N. Zou, and J. Liang, “Artificial lighting environment evaluation of the Japan Museum of Art based on the emotional response of observers,” Appl. Sci. 10(3), 1121 (2020). [CrossRef]
72. Y. Wang, Q. Liu, W. Gao, M. R. Pointer, Z. Huang, W. Chen, and J. Wu, “Interactive effect of illuminance and correlated colour temperature on colour preference and degree of white light sensation for chinese observers,” Optik 224, 165675 (2020). [CrossRef]
73. W. Davis and Y. Ohno, “Color quality scale,” Opt. Eng. 49(3), 033602 (2010). [CrossRef]
74. Y. Zhu, M. R. Luo, L. Xu, X. Liu, G. Cui, S. Fischer, P. Bodrogi, and T. Q. Khanh, “Investigation of memory colours across cultures,” in Proceedings of the 23th Color and Imaging Conference, (Society for Imaging Science and Technology (IS&T), Springfield, VA, USA, 2015), pp. 133–136.
75. Y. Zhu, M. R. Luo, S. Fischer, P. Bodrogi, and T. Q. Khanh, “The effectiveness of colour appearance attributes for enhancing image preference and naturalness,” in Proceedings of the 24th Color and Imaging Conference, (Society for Imaging Science and Technology (IS&T), Springfield, VA, USA, 2016), pp. 231–236.
76. Y. Zhu, M. R. Luo, S. Fischer, P. Bodrogi, and T. Q. Khanh, “Long-term memory color investigation: Culture effect and experimental setting factors,” J. Opt. Soc. Am. A 34(10), 1757–1768 (2017). [CrossRef]
77. S. Babilon, “On the color rendition of white light sources in relation to memory preference. Doctoral thesis: Technische Universität Darmstadt,” https://tuprints.ulb.tu-darmstadt.de/7799/ (2018). [Online; accessed 07-January-2021].
78. S. Babilon and T. Q. Khanh, “Color appearance rating of familiar real objects under immersive viewing conditions,” Color Res. Appl. 43, 551–568 (2018). [CrossRef]
79. S. Babilon and T. Q. Khanh, “A new metric for memory colour preference evaluation in lighting applications–Experiments, mathematical definition, and comparison with other colour rendering indices,” in Proceedings of the 29th Quadrennial Session of the CIE, (Commission Internationale de l’Éclairage (CIE), Vienna, Austria, 2019), pp. 251–260.
80. S. Babilon and T. Q. Khanh, “Impact of the adapted white point and the cultural background on memory color assessments,” Color Res. Appl. 45, 803–824 (2020). [CrossRef]
81. S. Babilon, J. Klabes, P. Myland, and T. Q. Khanh, “Memory colors and the assessment of color quality in lighting applications,” Opt. Express 29(18), 28968–28993 (2021). [CrossRef]
82. P. Bodrogi, S. Brückner, and T. Q. Khanh, “Ordinal scale based description of colour rendering,” Color Res. Appl. 36, 272–285 (2011). [CrossRef]
83. P. Bodrogi, S. Brückner, N. Krause, and T. Q. Khanh, “Semantic interpretation of color differences and color-rendering indices,” Color Res. Appl. 39, 252–262 (2014). [CrossRef]
84. S. A. Fotios and K. W. Houser, “Research methods to avoid bias in categorical ratings of brightness,” Leukos 5(3), 167–181 (2009). [CrossRef]
85. A. Žukauskas and M. S. Shur, “Color rendering metrics: Status, methods, and future development,” in Handbook of Advanced Lighting Technology, (Springer International Publishing, 2016), pp. 1–29.
86. K. Smet, J. Schanda, L. Whitehead, and R. Luo, “CRI2012: A proposal for updating the CIE colour rendering index,” Light. Res. Technol. 45(6), 689–709 (2013). [CrossRef]
87. K. Hashimoto, T. Yano, M. Shimizu, and Y. Nayatani, “New method for specifying color-rendering properties of light sources based on feeling of contrast,” Color Res. Appl. 32, 361–371 (2007). [CrossRef]
88. K. Smet and P. Hanselaer, “Memory and preferred colours and the colour rendition of white light sources,” Light. Res. Technol. 48(4), 393–411 (2016). [CrossRef]
89. A. David, P. T. Fini, K. W. Houser, Y. Ohno, M. P. Royer, K. A. G. Smet, M. Wei, and L. Whitehead, “Development of the IES method for evaluating the color rendition of light sources,” Opt. Express 23(12), 15888 (2015). [CrossRef]
90. W. Bao and M. Wei, “Change of gamut size for producing preferred color appearance from 20 to 15000 lux,” Leukos 17(1), 21–42 (2021). [CrossRef]
91. J. H. Lundeman, K. Herbst, M. Larsen, and L. Kessel, “Transmission measurements of human lenses,” in Therapeutic Laser Applications and Laser-Tissue Interactions IV, R. Sroka and L. D. Lilge, eds. (SPIE, 2009).
92. L. Kessel, J. H. Lundeman, K. Herbst, T. V. Andersen, and M. Larsen, “Age-related changes in the transmission properties of the human lens and their relevance to circadian entrainment,” J. Cataract. Refract. Surg. 36(2), 308–312 (2010). [CrossRef]
93. G. Banks, I. Heise, B. Starbuck, T. Osborne, L. Wisby, P. Potter, I. J. Jackson, R. G. Foster, S. N. Peirson, and P. M. Nolan, “Genetic background influences age-related decline in visual and nonvisual retinal responses, circadian rhythms, and sleep,” Neurobiol. Aging 36(1), 380–393 (2015). [CrossRef]
94. R. Feng, X. An-Cheng, and Z. Xi-Fang, “Change of the circadian effect of LED lighting with age,” Chin. J. Lumin. 37, 250–255 (2016). [CrossRef]
