< Back to previous page
Project
A New Metric for Evaluating the Colour Quality of White Light Sources with Reference to Memory Colours
Colour rendering, referring to the impact of a light source on the colour appearance of objects, has been investigated for many decades. Several metrics, not all successful, have been proposed in an attempt to quantify the effect of the light source. The current standard method, the colour rendering index Ra developed by the International Commission on Illumination (CIE), offers an objective measure of the colour rendering properties of a light source. However, such an objective measure may not be the best answer to the needs of lighting designers, architects and the people active in the shop and retail sector. End users and consumers are usually more interested in the more subjective aspects of colour rendering quality, i.e. in how appealing or how natural objects look. The failure of the CIE colour rendering index Ra to accurately assess the colour rendering quality of white light sources, especially narrowband sources such as Light-Emitting Diodes (LEDs), has made colour rendering a hot research topic.
The principal aim of this doctoral research project was to contribute to the international discussion on colour renderingby developing a new metric that meets most end-users expectations: a correct assessment of one or more of the subjective aspects of colour quality of a white light source, such as preference and attractiveness. In this work a new metric was developed that assesses the colour quality ofa light source with respect to the memory colours of a set of familiar objects. A memory colour approach was adopted because it corresponds more closely to the way people judge the colour quality of light sources ineveryday life than reference illuminant based approaches: nobody walks around with a reference source in their pockets to be able to judge the colour quality of the lighting in a room! The basic assumption of the approach is simple and intuitive: the better a light source renders objectcolours to what is expected, i.e. the memory colour, the better will bethe colour quality. Although memory colours have been investigated in the past, the available data was either insufficient for or inapplicable to the new memory colour quality metric. Therefore, new data was collected by investigating the memory colours of a set of ten familiar objects,as well as the human response to a deviation from these memory colours in a series of visual experiments. In the experiments the familiar objects were presented individually and illuminated by over one hundred different illumination spectra to a group of observers that were asked to rate the colour appearance of the objects with respect to what they thoughtthe objects look like in reality. The colour of the objects was changedby presenting them in a specially designed LED illumination box that created the illusion that the objects themselves changed colour and that kept the adaptation state of the observers constant. The good inter observer agreement suggested an average observer could be postulated by pooling the observer ratings for each familiar object. The pooled observer ratingswere modelled by a bivariate Gaussian function in IPT colour space.Based on the modelled functions a set of similarity functions were determined that describe the similarity of any object chromaticity with the memory colours of the familiar objects. As colour quality is assumed to be directly related to the similarity of an objects apparent colour with its memory colour, these similarity functions establish the basis of the memory colour quality metric. Practically, the colour quality of a white light source is determined as follows. For each familiar object a special memory colour quality indicator value is obtained by calculating the function value of the similarity function at the IPT object chromaticity under the test source. The object chromaticity is determined by first calculating the CIE 10° tristimulus values under the light source using the spectral reflectance of the object. Second, the tristimulus valuesare then converted to their corresponding colour using the CAT02 chromatic adaptation transform. The degree of chromatic adaptation is set depending on the illuminance. If unknown a value 0.90 corresponding to an illuminance of approximately 800 lx generally gives good results. Finally, the IPT object chromaticity is calculated from the corresponding tristimulus values. For a white light source, a general memory colour quality index is calculated as the geometric mean of the special indicator values. This number describes the overall agreement of the apparent object chromaticities with their memory colours and hence overall colour quality. Psychophysical experiments conducted in the framework of this doctoral research project as well as independent visual data collected from literature have validated the memory colour quality metric with high statistical significance as a metric with a high correlation with visual appreciation of light sources. It was shown to be the statistically best metric to assess the colour quality of a white light source in terms of visual appreciation among a large group of past and currently proposed metrics. However, the correlation with naturalness was rather low. The memory colour quality metric was also shown to be applicable to the design of new light sources with pleasing colour quality. Based on the optimization results of simulated tri- and tetrachromatic LED clusters a physicallamp was realized composed of red, green, blue LEDs and warm white phosphor LEDs. The superior colour quality of the lamp, optimized for the memory colour quality metric at 2700 K, was verified in a psychophysical rating experiment. It was found to perform significantly better than an incandescent lamp, of the same correlated colour temperature, in terms ofpreference, attractiveness and memory similarity. Furthermore, no statistical differences were observed in terms of naturalness and vividness.
Although there are still some potential shortcomings to the memory colour quality metric, such as the limited amount of test objects and the unknown influence of cultural differences, the version of the metric as it was developed during the doctoral research project has already shown great promise and could, even in its current form, be a significant contribution to the field of colour rendering. Anyway, this doctoral research project has made the international colour science community aware of the potential of a memory colour approach to colour quality evaluation ofwhite light sources.
The principal aim of this doctoral research project was to contribute to the international discussion on colour renderingby developing a new metric that meets most end-users expectations: a correct assessment of one or more of the subjective aspects of colour quality of a white light source, such as preference and attractiveness. In this work a new metric was developed that assesses the colour quality ofa light source with respect to the memory colours of a set of familiar objects. A memory colour approach was adopted because it corresponds more closely to the way people judge the colour quality of light sources ineveryday life than reference illuminant based approaches: nobody walks around with a reference source in their pockets to be able to judge the colour quality of the lighting in a room! The basic assumption of the approach is simple and intuitive: the better a light source renders objectcolours to what is expected, i.e. the memory colour, the better will bethe colour quality. Although memory colours have been investigated in the past, the available data was either insufficient for or inapplicable to the new memory colour quality metric. Therefore, new data was collected by investigating the memory colours of a set of ten familiar objects,as well as the human response to a deviation from these memory colours in a series of visual experiments. In the experiments the familiar objects were presented individually and illuminated by over one hundred different illumination spectra to a group of observers that were asked to rate the colour appearance of the objects with respect to what they thoughtthe objects look like in reality. The colour of the objects was changedby presenting them in a specially designed LED illumination box that created the illusion that the objects themselves changed colour and that kept the adaptation state of the observers constant. The good inter observer agreement suggested an average observer could be postulated by pooling the observer ratings for each familiar object. The pooled observer ratingswere modelled by a bivariate Gaussian function in IPT colour space.Based on the modelled functions a set of similarity functions were determined that describe the similarity of any object chromaticity with the memory colours of the familiar objects. As colour quality is assumed to be directly related to the similarity of an objects apparent colour with its memory colour, these similarity functions establish the basis of the memory colour quality metric. Practically, the colour quality of a white light source is determined as follows. For each familiar object a special memory colour quality indicator value is obtained by calculating the function value of the similarity function at the IPT object chromaticity under the test source. The object chromaticity is determined by first calculating the CIE 10° tristimulus values under the light source using the spectral reflectance of the object. Second, the tristimulus valuesare then converted to their corresponding colour using the CAT02 chromatic adaptation transform. The degree of chromatic adaptation is set depending on the illuminance. If unknown a value 0.90 corresponding to an illuminance of approximately 800 lx generally gives good results. Finally, the IPT object chromaticity is calculated from the corresponding tristimulus values. For a white light source, a general memory colour quality index is calculated as the geometric mean of the special indicator values. This number describes the overall agreement of the apparent object chromaticities with their memory colours and hence overall colour quality. Psychophysical experiments conducted in the framework of this doctoral research project as well as independent visual data collected from literature have validated the memory colour quality metric with high statistical significance as a metric with a high correlation with visual appreciation of light sources. It was shown to be the statistically best metric to assess the colour quality of a white light source in terms of visual appreciation among a large group of past and currently proposed metrics. However, the correlation with naturalness was rather low. The memory colour quality metric was also shown to be applicable to the design of new light sources with pleasing colour quality. Based on the optimization results of simulated tri- and tetrachromatic LED clusters a physicallamp was realized composed of red, green, blue LEDs and warm white phosphor LEDs. The superior colour quality of the lamp, optimized for the memory colour quality metric at 2700 K, was verified in a psychophysical rating experiment. It was found to perform significantly better than an incandescent lamp, of the same correlated colour temperature, in terms ofpreference, attractiveness and memory similarity. Furthermore, no statistical differences were observed in terms of naturalness and vividness.
Although there are still some potential shortcomings to the memory colour quality metric, such as the limited amount of test objects and the unknown influence of cultural differences, the version of the metric as it was developed during the doctoral research project has already shown great promise and could, even in its current form, be a significant contribution to the field of colour rendering. Anyway, this doctoral research project has made the international colour science community aware of the potential of a memory colour approach to colour quality evaluation ofwhite light sources.
Date:2 Jun 2008 → 20 Dec 2011
Keywords:Colour perception, LED, Colour rendering
Project type:PhD project