Color Research and Application最新文献

The paper describes a comprehensive test to evaluate the performance of current colour-difference models using available experimental datasets. In total, 28 individual datasets were accumulated to test 17 colour-difference formulae, 13 of them based on Uniform Colour Spaces (UCSs) in terms of the Standardized Residual Sum of Squares (STRESS) measure. The 28 datasets were divided into three groups: Large Colour-Difference data (LCD), Small Colour-Difference data for surface colours (SCDs), and Small Colour Difference data for display colours (SCDd). For each colour model, four versions were tested: the original model, and that including k_L-, Gamma- and k_L/Gamma, which are the lightness parametric factor, the colour-difference exponent factor, and the combination of both, respectively, optimized to fit particular dataset(s). The statistical F-test was applied to test the difference between each pair of models. Furthermore, parametric effects between the large/small colour-difference magnitudes, and between surface/display colours were investigated. The results showed that CAM16-UCS significantly outperformed the other models for all groups. It accurately predicted all types of data and should be proposed for colour-difference evaluation across all industries.

The Helmholtz-Kohlrausch (H-K) effect is investigated in relation to light colors of every hue, including those typical of print substrate colors that might be simulated on a graphic arts display. A method of adjustment is used in conjunction with a soft-proof setup, in which an achromatic stimulus is adjusted until it has the same lightness appearance as a set of test colors. Higher chroma colors are found to appear lighter than their metric $��L^{*}�$ would indicate. The H-K effect is found to be quite strong in bluish colors, but negligible in yellowish colors, consistent with several previous studies. However, qualitative analysis reveals a peak H-K effect in red-magenta hues. We propose a modification to Fairchild and Pirrotta's existing H-K lightness appearance function¹ which addresses the peak in red hues, and which may prove beneficial in hue-dependent applications.

Spectral reconstruction (SR) algorithms recover hyperspectral measurements from RGB camera responses. Statistical models at different levels of complexity are used to solve the SR problem—from the simplest closed-form regression, to sparse coding, to the complex deep neural networks (DNN). Recently, these methods were benchmarked based on the mean performance of the models and on a fixed set of real-world scenes, suggesting that more complex (more non-linear) models generally deliver better SR. In this paper, we investigate the relative performances of these models in terms of a real-world worst-case imaging condition called the Radiance Mondrian World (RMW) assumption. Under the RMW, testing hyperspectral images are composed of randomly-arranged and overlapping rectangular patches, where each patch is filled with one random radiance spectrum uniformly sampled from the convex closure of all natural radiances (i.e., all spectra in the concerned hyperspectral image dataset). Surprisingly, we show that all compared algorithms—regardless of their model complexity—degrade to broadly the same level of performance on our RMW testing set. Further, by retraining all models with an RMW training set, we show that increasing model complexity also does not help learning better SR mappings from the RMW images. That is, using simple regression is as good as using a DNN. This similarity of performance is also shown to hold for images adhering to the conventional Mondrian World assumption (random reflectances lit by a single, per scene, randomly selected light source).

This study explores the temporal changes in sentiment associated with eight color names over an 18-month period at four observation points. We focus on the valence aspect of sentiment. We collected four datasets, each separated by 6 months, and each containing 18 000 mentions of each of the eight color names in English from Twitter users around the world. We calculated the weighted average sentiment score of each instance when a color is mentioned. We find that purple and pink are the most positive in average sentiment score in all observation points, whereas brown, red, and orange are ranked as the lowest in average sentiment score. In terms of relative rank in sentiment value associated with the color names, we find the three consecutive datasets of July 2020, January 2021 and July 2021 are more consistent with one another, while the January 2022 dataset is more different from the earlier three datasets. This finding indicates that the temporal consistency in color-associated sentiment might maintain within 1 year, while evolve and show more difference in a longer timeline. This study is useful to marketing professionals by revealing that color names are associated with sentiment and that these associations can be monitored using social media data regularly. We suggest that marketers can use our method to analyse the color-associated sentiment of color names regularly, maybe on an annual basis, in order to choose color names wisely.

A psychophysical experiment was carried out to specify the preferred and natural memory colours corresponding to 24 familiar objects on mobile displays. These objects were selected to cover a large colour gamut in CIELAB space. One hundred and six observers from five different ethnic groups participated. Forty-nine colours for each object were rendered three-dimensionally to cover a large colour range. The results showed inter-observer variations are small between different ethnic groups, about 1.4 and 1.3 CIEDE2000 colour difference units for naturalness and preference, respectively. Comparing the preference and the naturalness colour centres, the preferred colours are brighter and more colourful than naturalness colours, by 1 and 2 units in L* and C*_ab attributes, respectively. Both sets of data show a great hue consistency. The results showed that different ethnic groups agreed well on the objects studied. It was found that the best-agreed objects are strawberry, green apple, tomato, corn and purple grape, and the worst-agreed objects are blueberry, broccoli, smurf, green pepper and blue sky.

Automated color measurement systems have become significant components of offset lithographic printing presses for better process control and reduced job waste. Some European press manufacturers are known to offer automated color measurement equipment that is confined to polarized density readings, which is uncommon in US. Introducing polarized densitometry into an environment where unpolarized densitometry is utilized can be problematic. Print professionals are on the lookout for a solution that is evidenced by a recent feature in a post from the IDEAlliance G7 Professionals forum seeking for a methodology to convert polarized to unpolarized readings. The present study applies the Bland–Altman method together with other statistical analyses to determine if a conversion between polarized and unpolarized densitometry can be established within a limited set of parameters. Three spectrodensitometers were used to record densitometric readings of wet and dry process inks on coated and uncoated drawdown proofs. The results showed that coated papers had significantly less variation between instruments compared to uncoated papers. Wet inks had more minor variations compared to dry inks. Furthermore, pure difference values and a Mann–Whitney U test indicated differences between unpolarized and polarized densitometric readings in all cases. Finally, t-tests performed on the resulting difference data for both unpolarized versus polarized and unpolarized versus predicted polarized that showed fixed bias by means of the computed p-values.

This work explores the potential of a compact hyperspectral camera, Specim IQ, for the remote colorimetric study of polychrome surfaces in controlled environments. These measurements are generally made with dedicated instruments, such as colorimeters or spectrophotometers, which require contact with the surface and coverage areas of the order of 10s of mm². These two characteristics, contact and a very small analysis area, can severely limit the study of polychrome surfaces, since the measured areas are not necessarily representative of the entire surface. In addition, it is not always possible to touch the objects being analyzed. A possible alternative is the use of compact hyperspectral cameras, such as Specim IQ, for the in situ study of the spectral and colorimetric characteristics of these surfaces. To better address this research, which is still in the preliminary phase, a 2 × 45°/0° geometry measuringement was used. The illumination of the image plane at 45° with respect to the camera made it possible to eliminate the components reflected specularly on the camera lens. With this shooting geometry, Labsphere Color and Gray Reflectance Standards (eight color and four gray standards) were analyzed and placed on seven different color backgrounds. With the spectral data acquired, it was possible to calculate the color of the targets and display the colorimetric values by means of three commonly used image processing software packages. In this way, it was possible to define for this hyperspectral camera a measurement-data processing procedure applicable to measurements in the laboratory aimed at studying the color of polychrome surfaces.

Exploring an efficient research method for understanding color preference is important to researchers and designers. This work compares four experimental methods for individual color preference research (N-alternative forced choice, rank-order, rating and paired comparison). Three psychophysical experiments were carried out with 338 participants. Participants were presented with six color patches (red, orange, yellow, green, blue and purple) arranged in a random order. This work suggested orange is the strongest preferred color and green is the weakest preferred using three individual color preference experimental methods with six hues. The Monte Carlo Analysis method further compares the result performance for four methods, which suggests the rating, paired comparison and rank-order methods are more stable than the N-alternative forced choice method when only a small number of participants take part in the experiment. For studies involving small numbers of participants (even less than 6), the rating, rank-order and pair comparison methods should be preferred.

This study mainly focused on exploring the multiple characteristics of specific associations from words to colors. Color association is bidirectional, representing the relationship from color to concept or concept to color. The association between word and color has multiple correspondences in both directions. Colors are associated with ideas, concepts, meanings, emotions, and so forth. However, it is often discussed as a single relationship from color to the concept. How the specific ideas or concepts associated with color, is rarely discussed. So, this work started from a specific concept (words) to colors and then investigated the characteristics of this association. A psychophysical experiment was carried out to collect the specific color associations from target words. Pearson correlation coefficient and K-means clustering methods were employed to analyze the associations from words to colors. Therefore, the main contribution of this study is to (1) indicate the stronger association from the word to color, the larger the color similarity of the related colors; (2) identify the specific characteristic of the association from a word to colors and the three-aspect characteristic model was summarized.