Evaluating colour models' performance between monitor and print images

The earlier experimental results (Color Res. Appl. 16, 166–180, 181–197; 18, 98–113, 191–209; 20, 18–28) have been further extended to include data obtained using complex images. Binocular memory and simultaneous matching techniques were used to assess the colour reproduction quality of displayed monitor images processed via eight colour models against a hardcopy (original) image illuminated in a viewing cabinet. The results from a panel of nine observers were used to compare different colour models' performance. It was found that the BFD chromatic adaptation transform outperformed the other models. The Hunt94 model, which gave a good fit to the earlier results, did not perform well. This indicates that there are differences in colour appearance between the complex and simple viewing fields. Other aspects were also investigated such as observer precision and repeatability, spatial uniformity of the monitor, image dependency, and the difference between the category judgment and paired comparison results. © 1996 John Wiley & Sons, Inc.     

Investigation of parametric effects using small colour differences

This experiment was carried out to investigate some viewing parameters affecting perceived colour differences. It was divided into eight phases. Each phase was conducted under a different set of experimental conditions including separations, neutral backgrounds, and psychophysical methods. Seventy‐five wool sample pairs were prepared corresponding to five CIE colour centers. The mean colour difference was three CIELAB units. Each pair was assessed by a panel of 21 observers using both the gray scale and pair comparison psychophysical methods. The assessments were carried out using the three different backgrounds (white, mid‐gray, and black) and a hairline gap between the samples. Assessments on the gray background were repeated using a large (3‐inch) gap between the samples. It was found that the visual results obtained from both psychophysical methods gave very similar results. The parametric effect was small, i.e., the largest effect was only 14% between the white and gray background conditions. These visual data were also used to test four colour‐difference formulae: CIELAB, CMC, BFD, and CIE94. The results showed that three advanced colour‐difference formulae performed much better than CIELAB. There was a good agreement between the current results and those from earlier studies. © 1999 John Wiley & Sons, Inc. Col Res Appl, 24, 331–343, 1999     

Investigation of colour appearance models for illumination changes across media

The key to achieving successful cross‐media colour reproduction is a reliable colour appearance model, which is capable of predicting the colour appearance across a variety of imaging devices under different viewing conditions. The two most commonly used media, CRT displays (soft copy) and printed images (hard copy), were included in this study using four complex images. The original printed images were captured using a digital camera and processed using eight colour appearance models (CIELAB, RLAB, LLAB, ATD, Hunt96, Nayatani97, CIECAM97s, and CAM97s2) and two chromatic adaptation transforms (von Kries and CMCCAT97). Psychophysical experiments were carried out to assess colour model performance in terms of colour fidelity by comparing soft‐copy and hard‐copy images. By employing the memory‐matching method, observers categorized the reproductions displayed on a CRT and compared them to the original printed images viewed in a viewing cabinet. The experiment was divided into three phases according to the different colour temperatures between the CRT and light source, i.e., print (D50, A, and A) and CRT (D93, D93, and D50), respectively). It was found that the CIECAM97s‐type models performed better than the other models. In addition, input parameters for each model had a distinct impact on model performance. © 2001 John Wiley & Sons, Inc. Col Res Appl, 26, 428–435, 2001     

A comprehensive model of colour appearance for related and unrelated colours of varying size viewed under mesopic to photopic conditions

CIE has recommended two previous appearance models, CIECAM97s and CIECAM02. However, these models are unable to predict the appearance of a comprehensive range of colours. The purpose of this study is to describe a new, comprehensive colour appearance model, which can be used to predict the appearance of colours under various viewing conditions that include a range of stimulus sizes, levels of illumination that range from scotopic through to photopic, and related and unrelated stimuli. In addition, the model has a uniform colour space that provides a colour‐difference formula in terms of colour appearance parameters. © 2016 Wiley Periodicals, Inc. Col Res Appl, 42, 293–304, 2017     

The ULAB colour space

A new colour space, named ULAB, is developed. It is derived from the CIELAB colour space and can be converted to and from CIELAB. Unlike modified CIELAB colour‐difference formulae, ULAB incorporates corrections for lightness, chroma, and hue differences into its colour coordinates. For the small magnitude colour difference data, it shows the performance as good as more complicated formulae such as CIEDE2000. ULAB shows another chance of developing a colour space approximately more uniform than CIELAB. © 2013 Wiley Periodicals, Inc. Col Res Appl, 40, 17–29, 2015     

A comparison of colour appearance for surface colours between outdoor and indoor environments

Psychophysical experiments of colour appearance, in terms of lightness, colourfulness, and hue, were conducted outdoors and indoors to investigate whether there was any difference in colour appearance between outdoor and indoor environments. A panel of 10 observers participated in the outdoor experiment, while 13 observers took part in the indoor experiment. The reference white had an average luminance of 12784 cd/m² in the outdoor experiment and 129 cd/m² in the indoor experiment. Test colours included 42 colour patches selected from the Practical Coordinate Color System to achieve a reasonable uniform distribution of samples in CIECAM02. Experimental results show that for both outdoor and indoor environments, there was good agreement between visual data and predicted values by CIECAM02 for the three colour appearance scales, with the coefficient of variation values all lower than 25 and the R² values all higher than 0.73, indicating little difference in the three dimensions of colour appearance between indoor and outdoor viewing conditions. Experimental data also suggest that the observers were more sensitive to variation in lightness for grayish colours than for highly saturated colours, a phenomenon that seems to relate with the Helmholtz-Kohlrausch effect. This phenomenon was modeled for predicting perceived lightness (J′) using the present experimental data. The new J′ model was tested using three extra sets of visual data obtained both outdoors and indoors, showing good predictive performance of the new model, with an average coefficient of variation of 14, an average R² of 0.88, and an average STRESS index of 14.18.     

An optical model relating (L*a*b*) values for a scattering surface covered with a scattering layer to (L*a*b*) values for the uncovered surface and its application to tooth colour

The appearance of human dentition is important both psychologically and commercially. Many people perceive the lightness and chromaticity of their teeth as key factors in their overall appearance leading to large businesses in materials for colour‐matched fillings and crowns and in tooth whitening products. The human eye is very sensitive to small colour differences, recognizing a row of highly colour‐matched crowns as unnatural yet seeing excessive colour variation or darkness as unattractive. One cause of tooth discolouration is a darkening of the dentine, visible through the enamel. This has lead the authors to develop a model capable of relating ( ) measurements on a scattering surface, in our case dentine, to ( ) measurements when overlaid by a translucent scattering layer, in our case tooth enamel. The model can be used when any scattering layer is superimposed on a coloured surface. In contrast to existing models, no spectral measurements are necessary allowing the use of colourimeters rather than spectrophotometers. However, there are limitations on the degree of colour saturation for both the coloured surface and the scattering layer as the model uses an approximation valid only for weakly saturated colours. As neither the enamel nor the dentine have strongly saturated colouration, the limitation is entirely acceptable for our work. The use of ( ) measurements directly rather than having to measure the spectrum of reflected light is of practical importance as such measurements in a dental surgery are impossible in all but exceptional cases whilst ( ) measurements in the surgery are routine. © 2014 Wiley Periodicals, Inc. Col Res Appl, 40, 504–517, 2015     

Unique hue data for colour appearance models. Part I: Loci of unique hues and hue uniformity

Psychophysical experiments were conducted to assess unique hues on a CRT display for a large sample of colour‐normal observers (n = 185). These data were then used to evaluate the most commonly used colour appearance model, CIECAM02, by transforming the CIEXYZ tristimulus values of the unique hues to the CIECAM02 colour appearance attributes, lightness, chroma and hue angle. We report two findings: (1) the hue angles derived from our unique hue data are inconsistent with the commonly used Natural Color System hues that are incorporated in the CIECAM02 model. We argue that our predicted unique hue angles (derived from our large dataset) provide a more reliable standard for colour management applications when the precise specification of these salient colours is important. (2) We test hue uniformity for CIECAM02 in all four unique hues and show significant disagreements for all hues, except for unique red which seems to be invariant under lightness changes. Our dataset is useful to improve the CIECAM02 model as it provides reliable data for benchmarking. © 2010 Wiley Periodicals, Inc. Col Res Appl, 2011     

Uniform colour spaces based on the DIN99 colour‐difference formula

Several colour‐difference formulas such as CMC, CIE94, and CIEDE2000 have been developed by modifying CIELAB. These formulas give much better fits for experimental data based on small colour differences than does CIELAB. None of these has an associated uniform colour space (UCS). The need for a UCS is demonstrated by the widespread use of the a*b* diagram despite the lack of uniformity. This article describes the development of formulas, with the same basic structure as the DIN99 formula, that predict the experimental data sets better than do the CMC and CIE94 colour‐difference formulas and only slightly worse than CIEDE2000 (which was optimized on the experimental data). However, these formulas all have an associated UCS. The spaces are similar in form to L*a*b*. © 2002 Wiley Periodicals, Inc. Col Res Appl, 27, 282–290, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.10066     

Prediction of Munsell appearance scales using various color‐appearance models

The chromaticities of the Munsell Renotation Dataset were applied to eight color‐appearance models. Models used were: CIELAB, Hunt, Nayatani, RLAB, LLAB, CIECAM97s, ZLAB, and IPT. Models were used to predict three appearance correlates of lightness, chroma, and hue. Model output of these appearance correlates were evaluated for their uniformity, in light of the constant perceptual nature of the Munsell Renotation data. Some background is provided on the experimental derivation of the Renotation Data, including the specific tasks performed by observers to evaluate a sample hue leaf for chroma uniformity. No particular model excelled at all metrics. In general, as might be expected, models derived from the Munsell System performed well. However, this was not universally the case, and some results, such as hue spacing and linearity, show interesting similarities between all models regardless of their derivation. © 2000 John Wiley & Sons, Inc. Col Res Appl, 25, 132–144, 2000     

Evaluation of Uniform Color Spaces Developed after the Adoption of CIELAB and CIELUV

Since the adoption of the color spaces CIELAB and CIELUV by the CIE in 1976, several other uniform spaces have been developed. We studied most of these spaces and evaluated their uniformity for small as well as larger color differences. Therefore, several criteria have been defined based on color discrimination data and appearance systems. The main difference between color spaces based on discrimination data and spaces that model appearance systems is reflected in a different size of the chroma distance unit compared with the lightness unit. If spaces based on the same kind of data (discrimination data or appearance systems) are compared with each other, they are all almost equally uniform. BFD (l:c), for example, is said to be more uniform than CMC(l:c), but, based on confidence intervals of 65%, there is no significant difference between them. If the proposed color difference formula of the CIE is compared with these distance functions, it also performs equally well. The SVF space and OSA 90 space on the other hand should be better than OSA 74. However, as opposed to what was expected, OSA 74 is slightly better; but, also in this case, the difference between the spaces is insignificant.     

Predicting visual similarity between colour palettes

This work is concerned with the prediction of visual colour difference between pairs of palettes. In this study, the palettes contained five colours arranged in a horizontal row. A total of 95 pairs of palettes were rated for visual difference by 20 participants. The colour difference between the palettes was predicted using two algorithms, each based on one of six colour-difference formulae. The best performance (r² = 0.86 and STRESS = 16.9) was obtained using the minimum colour-difference algorithm (MICDM) using the CIEDE2000 equation with a lightness weighing of 2. There was some evidence that the order (or arrangement) of the colours in the palettes was a factor affecting the visual colour differences although the MICDM algorithm does not take order into account. Application of this algorithm is intended for digital design workflows where colour palettes are generated automatically using machine learning and for comparing palettes obtained from psychophysical studies to explore, for example, the effect of culture, age, or gender on colour associations.     

Geometric relations between scales of small colour differences

Varying magnitude of colour differences from threshold up to moderate size in painted sample pairs at five CIE colour centers was estimated by grey scale assessment. Painted samples were produced for constant step width along the main axes of previously determined threshold (x,y,Y)‐ellipsoids with lightness variation at constant (x,y)‐chromaticity starting with threshold length and enlarging it five times for moderate magnitude of colour difference. Pairs were formed for linear extensions along axes and for diagonal combinations at equal step width between axes. The model under test assumes additive linear scale extension in constant proportions of the threshold (x,y,Y)‐ellipsoid for increasing magnitude of perceived colour difference and correlates perceptual main colour characters with main ellipsoid axes. Both assumptions were falsified to some degree: in general, magnitude of colour difference varies differently, though close to linear, and slightly subadditive for the three axes and for the different colour centers; the short (x,y)‐ellipse axis in some cases is not correlated with a perceptual hue vector component, and the main lightness direction sometimes is tilted in relation to the (x,y)‐plane. Three colour‐difference formulae do not provide better global predictions than the local (x,y,Y)‐ellipsoid formulae. The results may be used for more detailed modeling of colour‐difference formulae and for tolerance settings at different ranges of colour difference. © 1999 John Wiley & Sons, Inc. Col Res Appl, 24, 78–92, 1999     

A study of colour emotion and colour preference. Part II: Colour emotions for two‐colour combinations

Eleven colour‐emotion scales, warm–cool, heavy–light, modern–classical, clean–dirty, active–passive, hard–soft, harmonious–disharmonious, tense–relaxed, fresh–stale, masculine–feminine, and like–dislike, were investigated on 190 colour pairs with British and Chinese observers. Experimental results show that gender difference existed in masculine–feminine, whereas no significant cultural difference was found between British and Chinese observers. Three colour‐emotion factors were identified by the method of factor analysis and were labeled “colour activity,” “colour weight,” and “colour heat.” These factors were found similar to those extracted from the single colour emotions developed in Part I. This indicates a coherent framework of colour emotion factors for single colours and two‐colour combinations. An additivity relationship was found between single‐colour and colour‐combination emotions. This relationship predicts colour emotions for a colour pair by averaging the colour emotions of individual colours that generate the pair. However, it cannot be applied to colour preference prediction. By combining the additivity relationship with a single‐colour emotion model, such as those developed in Part I, a colour‐appearance‐based model was established for colour‐combination emotions. With this model one can predict colour emotions for a colour pair if colour‐appearance attributes of the component colours in that pair are known. © 2004 Wiley Periodicals, Inc. Col Res Appl, 29, 292–298, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.20024     

Proposed modification to the CIECAM02 colour appearance model to include the simultaneous contrast effects

A modified CIECAM02 colour appearance model, named CIECAM02‐m2, is proposed to enable CIECAM02 to predict the simultaneous contrast effect. The structure of the CIECAM02‐m2 is a development from CIECAM02, and contains two different procedures for modifying the reference white; one is for lightness and the other is for hue. The model was tested using a data set accumulated in this study and the LUTCHI data. The CV values for three colour attributes between predictions and experimental data were used to evaluate the accuracy of the model. The low CV values obtained show the performance of the CIECAM02‐m2 model to predict the simultaneous contrast effect satisfactorily. © 2007 Wiley Periodicals, Inc. Col Res Appl, 32, 121 – 129, 2007     

BFD (l:c) colour-difference formula Part 1ndashDevelopment of the formula

The available experimental data relating to small colour differences between pairs of surface colours have been combined together into sets including perceptibility results and acceptability results. Supplementary experiments have been carried out to enable all the previous visual results to be brought on to a common scale, and to provide extra information when this was considered necessary. A new colour-difference formula, BFD(l:c), has been developed using the combined experimental results. Various aspects of colour differences have been considered in turn to decide the form of formula required, and the constants in the formula have been optimised using the combined perceptibility and acceptability results. The new formula is similar in structure to the CMC(l:c) formula in most respects. However, it was found that a new term was required to take account of the fact that when chromaticity discrimination ellipses calculated from experimental results are plotted in a b space, they do not all point towards the neutral point. The experimental results were not very consistent with respect to possible tilting of discrimination ellipsoids relative to the xy plane. Overall it seems that any such tilting is quite small and in the direction implicit in the CMC and BFD formulae. Experimental results based on both acceptability and perceptibility judgements form part of the same overall pattern except for the weighting of lightness differences relative to hue and chroma differences.     

Evaluation of colour‐difference formulae for different colour‐difference magnitudes

Most of the colour‐difference formulae were developed to fit data sets having a limited range of colour‐difference magnitudes. Hence, their performances are uncertain when applying them to a range of colour differences from very small to very large colour differences. This article describes an experiment including three parts according to the colour‐difference magnitudes: large colour difference (LCD), small colour difference (SCD), and threshold colour difference (TCD) corresponding to mean ΔE values of 50.3, 3.5, and 0.6, respectively. Three visual assessment techniques were used: ratio judgement, pair comparison, and threshold for LCD, SCD, and TCD experiments, respectively. Three data sets were used to test six colour‐difference formulae and uniform colour spaces (CIELAB, CIE94, CIEDE2000, CAM02‐SCD, CAM02‐UCS, and CAM02‐LCD). The results showed that all formulae predicted visual results with great accuracy except CIELAB. CIEDE2000 worked effectively for the full range of colour differences, i.e., it performed the best for the TCD and SCD data and reasonably well for the LCD data. The three CIECAM02 based colour spaces gave quite satisfactory performance. © Wiley Periodicals, Inc. Col Res Appl, 2012     

Comparing large colour‐difference data sets

This study investigated the differences between different large colour‐difference (LCD) data sets (with a mean ΔE value about 10). Six data sets were studied. For each data set, various CIELAB based colour difference models were derived to fit the data. These models were compared to shed light on the difference between the different data sets. It was found that all data sets have very similar characteristics except for the Munsell data. Detailed investigation showed that the discrepancy is mainly due to the balance between the lightness and chromatic differences used previously for the Munsell data set. It was found that one unit of Munsell Value appears to be three times as large colour difference as one unit of Munsell Chroma at least under the experimental conditions for the data sets studied here. © 2010 Wiley Periodicals, Inc. Col Res Appl, 2011     

Colour size effect modelling

The visual phenomenon known as the colour size effect was investigated and two models were developed to predict the change in colour appearance of samples with six different sizes. The models are capable of transforming the colour appearance of a stimulus having a viewing field of 2° to that associated with a range of viewing fields. They are named the size effect correction and the size effect transform and are based on human perceptual attributes and human cone responses, respectively. The performance of both models was tested using the experimental data, and the results showed that the size effect transform performed better than the size effect correction. © 2010 Wiley Periodicals, Inc. Col Res Appl, 2012