CMC 2002 colour inconstancy index; CMCCON02

A colour inconstancy index, CMCCON02, is used for predicting the degree of colour inconstancy of a specimen defined by its spectral reflectance. CMCCON02 is different from the previously proposed CMCCON97 due to the replacement of the original chromatic adaptation transform (CMCCAT97) by CAT02. The latter is embedded in the CIE 2002 colour appearance model, CIECAM02. CAT02 is a simplified version of CMCCAT97 and gives more accurate predictions of the various experimental data sets. The CMCCON02 transform is in the process of being incorporated into ISO 105. This publication is sponsored by the Society's Colour Measurement Committee.     

Recommended colour-inconstancy index

One of the parameters of greatest interest in the design and production of coloured objects is their colour (in)constancy, i.e. their change in colour appearance with change in illuminant. Serious practical problems attend the visual assessment of colour inconstancy. Several instrumentally based models have been proposed but none has proved sufficiently reliable for routine use in the coloration industries. The Colour Measurement Committee (CMC) recommends a recently developed method, named the CMC 1997 colour-inconstancy index, which is based on a chromatic adaptation transform, CMCCAT97, that outperforms all others and is being increasingly adopted in international standards. This paper summarises the development of CMCCAT97 and the results of its testing, and gives details of the calculations necessary to CMCCAT97 and the new colour-inconstancy index. Publication sponsored by the Society's Colour Measurement Committee     

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     

Colour‐appearance modeling using feedforward networks with Bayesian regularization method. Part II: Reverse model

In Part I of this article, the development of a multilayer perceptrons feedforward artificial neural network model to predict colour appearance from colorimetric values was reported. Bayesian regularization was employed for the training of the network. In this part of the article, the reverse model, that is, the perdition of colorimetric values from the colour appearance attributes is reported using the same neural network design methodology developed in Part I. This study should contribute to the building of an artificial neural network–based colour appearance prediction, both forward and reverse, using the most comprehensive LUTCHI colour appearance data sets for training and testing. Good prediction accuracy and generalization ability were obtained using the neural networks built in the study. Because the neural network approach is of a black‐box type, colour appearance prediction using this method should be easier to apply in practice. © 2002 Wiley Periodicals, Inc. Col Res Appl, 27, 116–121, 2002; DOI 10.1002/col.10030     

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     

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     

Applying chromatic adaptation transforms to mixed adaptation conditions

A set of psychophysical experiments was conducted to investigate the state of adaptation between hardcopy and softcopy images when viewed under mixed illumination conditions. The performance of three chromatic adaptation transforms, CMCCAT97, CMCCAT2000 and CIECAT94, was compared to that of the S‐LMS, a mixed adaptation model. The adaptation ratios of each model were varied to investigate the state of adaptation of the human visual system. Printed complex images were used as originals. A series of softcopy reproduction pairs was displayed on a CRT with a D93 white point and compared to the original in an illuminated room. Three ambient lights were studied: a D50 simulator, an Illuminant A simulator and a Cool‐white fluorescence lamp. The experiments were divided into nine phases according to the changes of illuminants and the luminance levels. A simultaneous binocular matching technique was employed. Observers compared the hardcopy with a given pair of softcopies and identified which reproduction was the closer colour match. The results showed that the state of adaptation of the human visual system was between 40–60% adapted to the white point of the monitor regardless of the changes in illumination conditions. The overall results showed that CMCCAT2000 with proper adaptation ratio outperformed the other models and could be applied to mixed adaptation conditions. © 2003 Wiley Periodicals, Inc. Col Res Appl, 28, 436–444, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.10196     

Optimum solution of the CIECAM02 yellow–blue and purple problems

Brill [Color Res Appl 2006;31:142‐145] and Brill and Süsstrunk [Color Res Appl 2008;33:424‐426] found that CIECAM02 has the yellow–blue and purple problems and gave partial solutions to them. In this article we model the optimum solution to the yellow–blue and purple problems simultaneously as a constrained non‐linear optimization problem. An optimum solution resulting in a new CAT02 matrix is numerically obtained. This new matrix satisfies the nesting rule and performs better than the Hunt‐Pointer‐Estévez (HPE) matrix in predicting both corresponding colours and colour appearance data sets. Specifically, it was found that the new and HPE matrices performed significantly different on nine (out of 21) corresponding colour data sets and on all corresponding colours data sets as a whole. © 2014 Wiley Periodicals, Inc. Col Res Appl, 40, 491–503, 2015     

The CRI‐CAM02UCS colour rendering index

A new colour rendering index, CRI‐CAM02UCS, is proposed. It predicts visual results more accurately than the CIE CIR‐R_a. It includes two components necessary for predicting colour rendering in one metric: a chromatic adaptation transform and uniform colour space based on the CIE recommended colour appearance model, CIECAM02. The new index gave the same ranks as those of CIE‐R_a in the six lamps tested regardless the sample sets used. It was also found that the methods based on the size of colour gamut did not agree with those based on the test‐sample method. © 2011 Wiley Periodicals, Inc. Col Res Appl, 2012     

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.     

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     

Mathematical approach for predicting non‐negative tristimulus values using the CAT02 chromatic adaptation transform

It has been reported that for certain colour samples, the chromatic adaptation transform CAT02 imbedded in the CIECAM02 colour appearance model predicts corresponding colours with negative tristimulus values (TSVs), which can cause problems in certain applications. To overcome this problem, a mathematical approach is proposed for modifying CAT02. This approach combines a non‐negativity constraint for the TSVs of corresponding colours with the minimization of the colour differences between those values for the corresponding colours obtained by visual observations and the TSVs of the corresponding colours predicted by the model, which is a constrained non‐linear optimization problem. By solving the non‐linear optimization problem, a new matrix is found. The performance of the CAT02 transform with various matrices including the original CAT02 matrix, and the new matrix are tested using visual datasets and the optimum colours. Test results show that the CAT02 with the new matrix predicted corresponding colours without negative TSVs for all optimum colours and the colour matching functions of the two CIE standard observers under the test illuminants considered. However, the accuracy with the new matrix for predicting the visual data is approximately 1 CIELAB colour difference unit worse compared with the original CAT02. This indicates that accuracy has to be sacrificed to achieve the non‐negativity constraint for the TSVs of the corresponding colours. © 2011 Wiley Periodicals, Inc. Col Res Appl, 2011     

New colour appearance scales for describing saturation,vividness, blackness,and whiteness

This article describes the development of new models for predicting four colour appearance attributes: saturation, vividness, blackness, and whiteness. The new models were developed on the basis of experimental data accumulated in the authors' previous study, in which the four colour appearance attributes were scaled by 64 Korean and 68 British observers using the categorical judgment method. Two types of models were developed: the ellipsoid‐based and the hue‐based. For the former, the perceived saturation, vividness, blackness, and whiteness were modeled in the form of colour‐difference formulae between the test colour and a reference colour. For the latter, blackness, whiteness, and chromaticness scales were modeled by estimating hue‐dependent lightness and chroma values for the “full colour” in the framework of Adams' equation. The new models were tested using NCS data and were found to outperform some of the existing colour appearance models.     

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.     

Further improvements to CIECAM97s

The modifications to the CIECAM97s color appearance model suggested by Fairchild⁵ are reviewed in light of some recent experimental results, and some suggestions are made for refining and extending these modifications. © 2002 Wiley Periodicals, Inc. Col Res Appl, 27, 164–170, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.10047     

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     

An investigation of colour appearance for unrelated colours under photopic and mesopic vision

Data were obtained for the colour appearance of unrelated colours under photopic and mesopic conditions. The effects of changes in luminance level and stimulus size were investigated. The method used was magnitude scaling of brightness, colourfulness, and hue. Two stimulus sizes (10° and 0.5°) and four starting luminance levels (60, 5, 1, and 0.1, cd/m²) were used. The results at 0.1 cd/m² had large variations, so data were obtained for two additional stimulus sizes (1° and 2°) at this luminance level. Ten observers judged 50 unrelated colours. A total of 17,820 estimations were made. The observations were carried out in a completely darkened room, after 20 min adaptation; each test colour was presented on its own. Brightness and colourfulness were found to decrease with decreases of both luminance level and stimulus size. The CAM97u model predicted brightness more accurately than CIECAM02 but gave worse performance in predicting colorfulness. For hue, CAM97u and CIECAM02 both gave satisfactory predictions. Using the brightness correlate from CAM97u, a new colour‐appearance model based on CIECAM02 was developed specifically for unrelated colours under photopic and mesopic conditions, with parameters to allow for the effects of luminance level and stimulus size. © 2011 Wiley Periodicals, Inc. Col Res Appl, 2011;     

Quantifying colour appearance. part IV. Transmissive media

The experimental data from this study extends the LUTCHI colour-appearance data to cover transmissive media. Two further experiments were carried out: one used a large cut-sheet transparency viewed using a back-lit illuminator, and another used a 35-mm slide projected onto a white screen. These new data were used to reveal the changes in colour appearance caused by different viewing parameters studied, and to evaluate the predictive accuracy of five uniform colour spaces and colour-appearance models. The results show that Hunt's 91 model (developed to fit earlier experimental results) did not perform as well as it did in the previous studies using nontransmissive media. This implies that there are large differences in perceived colour appearance between transmissive and nontransmissive media viewing conditions. Some modifications were subsequently made to Hunt's 91 model and the predictive accuracy was greatly improved.     

Illuminant estimation for object recognition

Comparing colour histograms of images has been shown to be a powerful technique for discriminating among large sets of images. However, these histograms depend not only on the properties of imaged objects but also on the illumination under which the objects are captured. If this illumination dependence is not accounted for prior to constructing the colour histograms, colour‐based image indexing will fail when illumination changes. This failure can be addressed by correcting the RGBs in an image to corresponding RGBs representing the same scene but under a standard reference illuminant prior to constructing the histograms. To perform this correction of RGBs, it is necessary to have a measurement or, more commonly, an estimate of the illumination in the original scene. Many authors have proposed illuminant estimation (or colour constancy) algorithms to obtain such an estimate. Unfortunately, the results of colour histogram matching experiments under varying illumination conditions have shown that existing estimation algorithms do not provide a sufficiently good estimate of the scene illuminant to enable this approach to work. In this article we report on the results of our repetition of those experiments, but this time using a new illuminant estimation algorithm—the so‐called color by correlation approach, which has been shown to afford significantly better performance than previous algorithms. The results of this new experiment show that when this new algorithm is used to preprocess images, a significant improvement in colour histogram matching performance is achieved. Indeed, performance is close to the theoretically optimal level of performance, that is, close to that which can be obtained using actual measurements of the scene illumination. © 2002 Wiley Periodicals, Inc. Col Res Appl, 27, 260–270, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.10064