Predicting Facial Attractiveness from Colour Cues: A New Analytic Framework.

Various facial colour cues were identified as valid predictors of facial attractiveness, yet the conventional univariate approach has simplified the complex nature of attractiveness judgement for real human faces. Predicting attractiveness from colour cues is difficult due to the high number of candidate variables and their inherent correlations. Using datasets from Chinese subjects, this study proposed a novel analytic framework for modelling attractiveness from various colour characteristics. One hundred images of real human faces were used in experiments and an extensive set of 65 colour features were extracted. Two separate attractiveness evaluation sets of data were collected through psychophysical experiments in the UK and China as training and testing datasets, respectively. Eight multivariate regression strategies were compared for their predictive accuracy and simplicity. The proposed methodology achieved a comprehensive assessment of diverse facial colour features and their role in attractiveness judgements of real faces; improved the predictive accuracy (the best-fit model achieved an out-of-sample accuracy of 0.66 on a 7-point scale) and significantly mitigated the issue of model overfitting; and effectively simplified the model and identified the most important colour features. It can serve as a useful and repeatable analytic tool for future research on facial impression modelling using high-dimensional datasets.

Domain and range of the CIECAM16 forward transformation.

The domain and range of the CIECAM16 forward transformation was numerically determined and visualized for CIE standard illuminants, using a linear programming approach that provides the gamuts and colour solids for optimum colours. The effect of the surround, adapting luminance, and luminance of the background on the range of the CIECAM16 forward transformation were individually analyzed, showing that their ranges increased when the surround changed from dark to dim or average, the adapting luminance increased, or the luminance of the background decreased. The proposed methodology for the determination and visualization of the domain and range of the CIECAM16 forward transformation can be used for any illuminant, as well as for CIECAM02, CAM16, CAM02-UCS and CAM16-UCS. The findings of this paper not only solve the long-term unresolved domain and range problems of the CIE colour appearance models, but also find applications in cross-media colour reproduction. Furthermore, it was also found that some non-CIE colours are inside the International Color Consortium Profile Connection Space (ICC PCS), and some CIE colours are not included in that space.

Research on the interactive perception of color paper-cutting culture rendering based on AR.

To address the digital medium preference of the millennial generation, this study utilizes augmented reality (AR) technology for rendering color paper cuttings by developing a mobile terminal based on Unity 3D and 3ds Max to demonstrate digital works scanned by paper-cutting entities. Three subjective scaling experiments are conducted to evaluate the aesthetics, viewership, and impression of four genres of digital color paper cutting. The results show that observers have more preference for warm background with the maximum scaling value at the 7.5 mm/s playback speed and a specific superimposed order. Importantly, current experimental design and interactive evaluation provide a reference for AR display parameters.

Spectra Reconstruction for Human Facial Color from RGB Images via Clusters in 3D Uniform CIELab* and Its Subordinate Color Space.

Previous research has demonstrated the potential to reconstruct human facial skin spectra based on the responses of RGB cameras to achieve high-fidelity color reproduction of human facial skin in various industrial applications. Nonetheless, the level of precision is still expected to improve. Inspired by the asymmetricity of human facial skin color in the CIELab* color space, we propose a practical framework, HPCAPR, for skin facial reflectance reconstruction based on calibrated datasets which reconstruct the facial spectra in subsets derived from clustering techniques in several spectrometric and colorimetric spaces, i.e., the spectral reflectance space, Principal Component (PC) space, CIELab*, and its three 2D subordinate color spaces, La*, Lb*, and ab*. The spectra reconstruction algorithm is optimized by combining state-of-art algorithms and thoroughly scanning the parameters. The results show that the hybrid of PCA and RGB polynomial regression algorithm with 3PCs plus 1st-order polynomial extension gives the best results. The performance can be improved substantially by operating the spectral reconstruction framework within the subset classified in the La* color subspace. Comparing with not conducting the clustering technique, it attains values of 25.2% and 57.1% for the median and maximum errors for the best cluster, respectively; for the worst, the maximum error was reduced by 42.2%.

Interim connection space based on colorimetric values for spectral image compression and reproduction.

A new interim and connection space (ICS) and its reconstruction method are proposed. The proposed ICS,tD65A, consists of six colorimetric values or two sets of tristimulus values under CIE illuminant D65 and A respectively. In addition, a new spectral decomposition based on the tD65A ICS and the Wiener Estimation matrix MW was introduced for an improved spectral reconstruction. Accompanying the tD65A ICS, m important basis vectors for the metameric black space based on the new spectral decomposition, and a mapping matrix MP,k via a polynomial model of order k, were trained so that both the spectral and colorimetric accuracies for the reconstructed reflectance can be further enhanced. The proposed ICS and its reconstruction method can ensure exact colorimetric matches under two (real rather than synthetic) illuminants D65 and A, which is an advantage compared with other ICSs. The performance of the proposed method was tested and compared with five other ICSs using the NCS dataset and three spectral images respectively, using RMSE and GFC to measure the spectral accuracy, and using CIEDE2000 colour differences to measure the colorimetric accuracy under three types of illuminants (continuous, fluorescent, and LED). Performance test results showed the proposed methods outperform other ICSs in terms of both spectral accuracy and colorimetric measures (RMSE, GFC, and CIEDE2000 colour difference). Therefore, it is expected the proposed ICS and its reconstruction method can play an important role in spectral image compression and reproduction applications.

Influences of shape, size, and gloss on the perceived color difference of 3D printed objects.

In order to study the influence and mechanisms of color differences using 3D-shaped objects, 440 pairs of 3D samples surrounding five CIE color centers (gray, red, yellow, green, and blue) with the variations of gloss, size, and shape were prepared by a Sailner 3D color printer, and their color differences were assessed by 26∼45 observers using the gray-scale method. The new color difference data were used to investigate the parametric effects (gloss, 3D shape, and size) on the perceived color difference. Results indicate that, for 3D objects, high gloss and small size objects (2 cm) raise smaller visual color differences than matte and large size objects (4 cm), and the visual color difference of spheres is larger than that of the cone and cylinder sample pairs. The chromaticity ellipses indicated that the glossy samples with different shapes will arouse fairly different visual perceptions, especially for sphere and cylinder samples.

Optimizing Color-Difference Formulas for 3D-Printed Objects.

Based on previous visual assessments of 440 color pairs of 3D-printed samples, we tested the performance of eight color-difference formulas (CIELAB, CIEDE2000, CAM02-LCD, CAM02-SCD, CAM02-UCS, CAM16-LCD, CAM16-SCD, and CAM16-UCS) using the standardized residual sum of squares (STRESS) index. For the whole set of 440 color pairs, the introduction of kL (lightness parametric factor), b (exponent in total color difference), and kL + b produced an average STRESS decrease of 2.6%, 26.9%, and 29.6%, respectively. In most cases, the CIELAB formula was significantly worse statistically than the remaining seven formulas, for which no statistically significant differences were found. Therefore, based on visual results using 3D-object colors with the specific shape, size, gloss, and magnitude of color differences considered here, we concluded that the CIEDE2000, CAM02-, and CAM16-based formulas were equivalent and thus cannot recommend only one of them. Disregarding CIELAB, the average STRESS decreases in the kL + b-optimized formulas from changes in each one of the four analyzed parametric factors were not statistically significant and had the following values: 6.2 units changing from color pairs with less to more than 5.0 CIELAB units; 2.9 units changing the shape of the samples (lowest STRESS values for cylinders); 0.7 units changing from nearly-matte to high-gloss samples; and 0.5 units changing from 4 cm to 2 cm samples.

Color-difference evaluation for 3D objects.

A psychophysical experiment using 3D printed samples was conducted to investigate the change of perceived color differences caused by two different illuminations and two 3D sample shapes. 150 pairs of 3D printed samples around five CIE color centers [Color Res. Appl. 20, 399-403, 1995], consisting of 75 pairs of spherical samples and 75 pairs of flat samples, with a wide range of color differences covering from small to large magnitude, were printed by an Mcor Iris paper-based 3D color printer. Each pair was assessed twice by a panel of 10 observers using a gray-scale psychophysical method in a spectral tunable LED viewing cabinet with two types of light sources: diffuse lighting with and without an additional overhead spotlight. The experimental results confirmed that the lighting conditions had more effect on the perceived color difference between complex 3D shapes than between 2D objects. The results for 3D and 2D objects were more similar under only diffuse lighting. Current 3D results had good correlations with previous ones [Color Res. Appl. 24, 356-368, 1999; J. Opt. Soc. Am. A 36, 789-799, 2019] using 2D samples with large color differences, meaning that color-difference magnitude had more effect on perceived color differences than sample shape and lighting. Considering ten modern color-difference formulas, the best predictions of the current experimental data were found for CAM02-LCD formula [Color Res. Appl. 31, 320-330, 2006]. For current results, it was also found that predictions of current color-difference formulas were below average inter-observer variability, and remarkable improvements were found by adding power corrections [Opt. Express 23, 597-610, 2015].

Spectral Reconstruction Using an Iteratively Reweighted Regulated Model from Two Illumination Camera Responses.

An improved spectral reflectance estimation method was developed to transform captured RGB images to spectral reflectance. The novelty of our method is an iteratively reweighted regulated model that combines polynomial expansion signals, which was developed for spectral reflectance estimation, and a cross-polarized imaging system, which is used to eliminate glare and specular highlights. Two RGB images are captured under two illumination conditions. The method was tested using ColorChecker charts. The results demonstrate that the proposed method could make a significant improvement of the accuracy in both spectral and colorimetric: it can achieve 23.8% improved accuracy in mean CIEDE2000 color difference, while it achieves 24.6% improved accuracy in RMS error compared with classic regularized least squares (RLS) method. The proposed method is sufficiently accurate in predicting the spectral properties and their performance within an acceptable range, i.e., typical customer tolerance of less than 3 DE units in the graphic arts industry.

Investigating unique hues at different chroma levels with a smaller hue angle step.

Unique hue plays a critical role in color appearance models and uniform color spaces. Past studies investigating unique hues commonly used 40 Munsell samples with the same chroma and lightness levels to produce color stimuli, with a hue angle step of 9°. These 40 samples were always simultaneously presented to the observers. Both the larger hue angle step and the simultaneous presentation of the samples may help to reduce the variations. In this study, we reduced the hue angle step to 5° and each stimulus was individually presented to the observer, which resulted in larger inter- and intra-observer variations. The results suggested that the hue angles of the unique hues in both CIECAM02 and CIELAB should be revised, but both CIECAM02 and CIELAB had good hue uniformity at the hue angles of the four unique hues.

Spectra estimation from raw camera responses based on adaptive local-weighted linear regression.

An improved spectral reflectance estimation method is developed to transform raw camera RGB responses to spectral reflectance. The novelty of our method is to apply a local weighted linear regression model for spectral reflectance estimation and construct the weighting matrix using a Gaussian function in CIELAB uniform color space. The proposed method was tested using both a standard color chart and a set of textile samples, with a digital RGB camera and by ten times ten-fold cross-validation. The results demonstrate that our method gives the best accuracy in estimating both the spectral reflectance and the colorimetric values in comparison with existing methods.

Chromatic and luminance sensitivity for skin and skinlike textures.

Despite the importance of the appearance of human skin for theoretical and practical purposes, little is known about visual sensitivity to subtle skin-tone changes, and whether the human visual system is indeed optimized to discern skin-color changes that confer some evolutionary advantage. Here, we report discrimination thresholds in a three-dimensional chromatic-luminance color space for natural skin and skinlike textures, and compare these to thresholds for uniform stimuli of the same mean color. We find no evidence that discrimination performance is superior along evolutionarily relevant color directions. Instead, discriminability is primarily determined by the prevailing illumination, and discrimination ellipses are aligned with the daylight locus. More specifically, the area and orientation of discrimination ellipses are governed by the chromatic distance between the stimulus and the illumination. Since this is true for both uniform and textured stimuli, it is likely to be driven by adaptation to mean stimulus color. Natural skin texture itself does not confer any advantage for discrimination performance. Furthermore, we find that discrimination boundaries for skin, skinlike, and scrambled skin stimuli are consistently larger than those for uniform stimuli, suggesting a possible adaptation to higher order color statistics of skin. This is in line with findings by Hansen, Giesel, and Gegenfurtner (2008) for other natural stimuli (fruit and vegetables). Human observers are also more sensitive to skin-color changes under simulated daylight as opposed to fluorescent light. The reduced sensitivity is driven by a decline in sensitivity along the luminance axis, which is qualitatively consistent with predictions from a Von Kries adaptation model.

Theoretical consideration on convergence of the fixed-point iteration method in CIE mesopic photometry system MES2.

Currently the fixed-point iteration method with initial guess m0=0.5 is officially recommended by the CIE MES2 system [CIE 191:2010] in order to compute the adaptation coefficient m and the mesopic luminance L m e s . However, recently, Gao et al. [Opt. Express25, 18365 (2017)] and Shpak et al. [Lighting Res. Technol.49, 111 (2017)] have numerically found that the fixed-point iteration method could be not convergent for large values of S/P. Shpak et al. suspected that, to achieve convergence, the S/P ratio cannot be greater than 17. In this paper, a theoretical consideration for the CIE MES2 system is given. Namely, it is shown that the ratio S/P be smaller than a constant C2 (≈18.1834) is a sufficient condition for the convergence of the fixed-point iteration method. In addition, a new initial guess strategy, achieving faster convergence, is proposed.

Gamut Volume Index: a color preference metric based on meta-analysis and optimized colour samples.

A novel metric named Gamut Volume Index (GVI) is proposed for evaluating the colour preference of lighting. This metric is based on the absolute gamut volume of optimized colour samples. The optimal colour set of the proposed metric was obtained by optimizing the weighted average correlation between the metric predictions and the subjective ratings for 8 psychophysical studies. The performance of 20 typical colour metrics was also investigated, which included colour difference based metrics, gamut based metrics, memory based metrics as well as combined metrics. It was found that the proposed GVI outperformed the existing counterparts, especially for the conditions where correlated colour temperatures differed.

Improved computation of the adaptation coefficient in the CIE system of mesopic photometry.

New values of parameters a and b are proposed for the CIE system of mesopic photometry MES2 [CIE Publication 191:2010], because from the original values this model may have no solution or multi-solutions. From the new values of parameters a and b it is shown that the CIE MES2 system has a unique solution. The difference however, between the original and the new values of parameters a and b is very small and the changes do not affect previous conclusions based on the MES2 model. To compute such a solution, we propose a Bisection-Newton method which exhibits fast convergence (8 iterations in the worst case), and improves the fixed-point method recommended by the CIE MES2 system, which has convergence problems for high values of the photopic luminance and very high values of the scotopic/photopic ratio. Comparative results for the fixed-point method, the Bisection method, the Newton method, and the Bisection-Newton method, in terms of the number of iterations necessary for convergence and the computation time used, are reported.

Improved method for skin reflectance reconstruction from camera images.

A improved spectral reflectance reconstruction method is developed to transform camera RGB to spectral reflectance for skin images. Rather than using conventional direct or two-step processes, we transform camera RGB to skin reflectance directly using a principal component analysis (PCA) approach. The novelty in our direct method (RGB to spectra) is the use of a skin-specific colour characterisation chart with spectra closer to human skin spectra, and a new database of skin reflectances to derive the PCA bases. The experimental results using the facial images of 17 subjects demonstrate that our new direct method gives a significantly better performance than conventional, two-step methods and direct methods with traditional characterization charts. This new spectral reconstruction algorithm is sufficiently precise to reconstruct spectral properites relating to chromophores and its performance is within the acceptable range for maxillofacial soft tissue prostheses (error < 3 ΔE*ab units).

Accurate method for computing correlated color temperature.

For the correlated color temperature (CCT) of a light source to be estimated, a nonlinear optimization problem must be solved. In all previous methods available to compute CCT, the objective function has only been approximated, and their predictions have achieved limited accuracy. For example, different unacceptable CCT values have been predicted for light sources located on the same isotemperature line. In this paper, we propose to compute CCT using the Newton method, which requires the first and second derivatives of the objective function. Following the current recommendation by the International Commission on Illumination (CIE) for the computation of tristimulus values (summations at 1 nm steps from 360 nm to 830 nm), the objective function and its first and second derivatives are explicitly given and used in our computations. Comprehensive tests demonstrate that the proposed method, together with an initial estimation of CCT using Robertson's method [J. Opt. Soc. Am. 58, 1528-1535 (1968)], gives highly accurate predictions below 0.0012 K for light sources with CCTs ranging from 500 K to 106 K.

The achromatic locus: effect of navigation direction in color space.

An achromatic stimulus is defined as a patch of light that is devoid of any hue. This is usually achieved by asking observers to adjust the stimulus such that it looks neither red nor green and at the same time neither yellow nor blue. Despite the theoretical and practical importance of the achromatic locus, little is known about the variability in these settings. The main purpose of the current study was to evaluate whether achromatic settings were dependent on the task of the observers, namely the navigation direction in color space. Observers could either adjust the test patch along the two chromatic axes in the CIE u*v* diagram or, alternatively, navigate along the unique-hue lines. Our main result is that the navigation method affects the reliability of these achromatic settings. Observers are able to make more reliable achromatic settings when adjusting the test patch along the directions defined by the four unique hues as opposed to navigating along the main axes in the commonly used CIE u*v* chromaticity plane. This result holds across different ambient viewing conditions (Dark, Daylight, Cool White Fluorescent) and different test luminance levels (5, 20, and 50 cd/m(2)). The reduced variability in the achromatic settings is consistent with the idea that internal color representations are more aligned with the unique-hue lines than the u* and v* axes.

Digital Grading the Color Fastness to Rubbing of Fabrics Based on Spectral Reconstruction and BP Neural Network.

To digital grade the staining color fastness of fabrics after rubbing, an automatic grading method based on spectral reconstruction technology and BP neural network was proposed. Firstly, the modeling samples are prepared by rubbing the fabrics according to the ISO standard of 105-X12. Then, to comply with visual rating standards for color fastness, the modeling samples are professionally graded to obtain the visual rating result. After that, a digital camera is used to capture digital images of the modeling samples inside a closed and uniform lighting box, and the color data values of the modeling samples are obtained through spectral reconstruction technology. Finally, the color fastness prediction model for rubbing was constructed using the modeling samples data and BP neural network. The color fastness level of the testing samples was predicted using the prediction model, and the prediction results were compared with the existing color difference conversion method and gray scale difference method based on the five-fold cross-validation strategy. Experiments show that the prediction model of fabric color fastness can be better constructed using the BP neural network. The overall performance of the method is better than the color difference conversion method and the gray scale difference method. It can be seen that the digital rating method of fabric staining color fastness to rubbing based on spectral reconstruction and BP neural network has high consistency with the visual evaluation, which will help for the automatic color fastness grading.

Blue-green color categorization in Mandarin-English speakers.

Observers are faster to detect a target among a set of distracters if the targets and distracters come from different color categories. This cross-boundary advantage seems to be limited to the right visual field, which is consistent with the dominance of the left hemisphere for language processing [Gilbert et al., Proc. Natl. Acad. Sci. USA 103, 489 (2006)]. Here we study whether a similar visual field advantage is found in the color identification task in speakers of Mandarin, a language that uses a logographic system. Forty late Mandarin-English bilinguals performed a blue-green color categorization task, in a blocked design, in their first language (L1: Mandarin) or second language (L2: English). Eleven color singletons ranging from blue to green were presented for 160 ms, randomly in the left visual field (LVF) or right visual field (RVF). Color boundary and reaction times (RTs) at the color boundary were estimated in L1 and L2, for both visual fields. We found that the color boundary did not differ between the languages; RTs at the color boundary, however, were on average more than 100 ms shorter in the English compared to the Mandarin sessions, but only when the stimuli were presented in the RVF. The finding may be explained by the script nature of the two languages: Mandarin logographic characters are analyzed visuospatially in the right hemisphere, which conceivably facilitates identification of color presented to the LVF.