Research and application of color features of female court attire in Tang Dynasty based on the COLORO system

Abstract

The female court attire during the Tang Dynasty was splendid in color and worthy of research and reference. In this study, this type of attire was categorized according to social status, occasion, and occupation. The color features of each attire were extracted using a clustering algorithm and analyzed with the COLORO system. The results showed that the color schemes of female court attire were closely tied to social status and occasion. Noblewomen favored red tones, while the maidservants’ attire was dominated by dark blue or brown hues. The color schemes also adhered to the principle of “coloring according to the master or occasion.” Additionally, a ChatGPT model was utilized to generate innovative color schemes. Three evaluation metrics, color harmony, artistic appeal, and historical appropriateness, were applied to assess the color schemes. Ultimately, this color-matching method has proven to be effective. These conclusions not only improve the theoretical system of clothing color schemes during the Tang dynasty but also provide a digital application pathway for contemporary Hanfu design.

Keywords

Tang dynasty female court attire color matching cluster analysis COLORO system Hanfu design

Introduction

The female court attires from the Tang Dynasty, characterized by its rich and vibrant color schemes and striking visual expressions, stands as a hallmark of the artistry of color in ancient Chinese clothing.^1,2 By systematically exploring its color composition, deconstructing the principles behind its color coordination, and innovatively applying traditional aesthetic values to the design of contemporary Hanfu (traditional Chinese costumes), we not only imbue modern clothing with profound historical and cultural significance but also enable contemporary Hanfu design to overcome the constraints of homogenization. This approach facilitates the creation of an original design language that integrates both traditional cultural heritage and the spirit of the times, thereby offering substantial practical value in enhancing the cultural identity and artistic appeal of Hanfu.

Currently, research on the color matching of Tang Dynasty women’s clothing is largely limited to qualitative descriptions and has not yet developed into a systematic and in-depth research framework. Dai and Wang investigated the color matching characteristics of women’s skirts from the Tang Dynasty period and found that they were primarily composed of red hues with varying purity and brightness.³ The study by Fan revealed that red was the most frequently used color in women’s clothing during the Tang Dynasty based on an examination of numerous portraits of women from this period.⁴ From the perspective of color aesthetics, a study by Zhang pointed out that Tang Dynasty women’s clothing featured bright and vivid tones, characterized by strong color contrasts and remarkable visual tension.⁵ Similarly, Tian conducted a study focusing on the color characteristics of women’s robes and skirts from the Tang Dynasty, analyzing the matching patterns of clothing colors and their underlying causes.⁶ However, an examination of existing research reveals a lack of analysis on clothing color quantification analysis. Studies on key quantitative indicators, such as the distribution of colors, area ratios of colors, and the frequency of primary and secondary color combinations, remain insufficient. Furthermore, in-depth exploration and theoretical refinement of color matching principles, such as hue composition, brightness levels, and chroma variation, are rare and need further development.

In addition, current research on the color evolution characteristics of Tang Dynasty clothing primarily unfolds along a chronological timeline. Na⁷ and Wei et al.,⁸ through their analyses of Dunhuang frescoes, have outlined the emotional characteristics and evolutionary patterns of women’s clothing colors during the early, flourishing, and middle-to-late periods of the Tang Dynasty, respectively. Their studies found that the colors of clothing in the early periods of the Tang Dynasty were characterized by a delicate and soft emotional tone. During the flourishing period of the Tang Dynasty, clothing displayed a luxurious and splendid style, while the colors of clothing in the middle-to-late periods of the Tang Dynasty became subdued and gentle. Meanwhile, Wei et al. pointed out that the color evolution characteristics of clothing were closely related to the social and economic changes during the Tang Dynasty. However, beyond historical contexts, factors such as social class and occupational roles also significantly influenced the color variations of clothing in the Tang Dynasty, which should not be overlooked. Currently, most studies of this nature are qualitatively described based on historical texts from the Tang Dynasty, while quantitative analyses of clothing illustrations from this period remain insufficient and require further in-depth exploration.

Based on the previously mentioned research status, this study constructs a four-tier research framework which consists of color extraction, feature analysis, creative transformation, and comprehensive evaluation. Firstly, a vast number of portrait paintings from the Tang Dynasty were systematically collected. A detailed classification of the female court attire during the Tang Dynasty was conducted, examining it from three dimensions: social class, occasion, and occupation. Secondly, a K-means clustering algorithm was applied to analyze the color-matching features of various types of clothing. Meanwhile, the COLORO system was introduced to analyze in depth the intrinsic patterns of hue distribution, value gradients, and chroma variation from multiple dimensions. Thirdly, the color-matching rules extracted from the study were deeply integrated into the ChatGPT model, leveraging big data to facilitate the rapid generation of color schemes for Tang-style Hanfu. Finally, the color harmony, aesthetic appeal, and historical appropriateness of color schemes were assessed based on feedback from subjects including researchers of Chinese traditional clothing, Hanfu designers, and Hanfu enthusiasts. The aim of this research was to analyze the color-matching rules of female court attire from the Tang Dynasty and to apply these rules to contemporary Hanfu design. This study will explore the modern transformation of traditional color systems, enhance the cultural depth and artistic value of original Hanfu designs, and promote both the living inheritance and innovative expression of clothing culture from the Tang Dynasty in contemporary Hanfu.

Methods

Paintings and Samples

Fifteen classical court paintings from the Tang Dynasty were selected as research materials for this study, as shown in Figure 1. All materials related to Tang Dynasty court paintings are sourced from publicly accessible online digital resources. During the screening process, priority was given to high-resolution scanned copies, versions reprinted on academic platforms, and those with explicitly labeled original collection information. The original collecting institution and website index for each painting have been documented in Table 1, ensuring the authenticity and academic validity of the images. All materials in this study are used exclusively for illustrative analysis and to support arguments in the thesis, with no commercial application involved.

Figure 1.

Fifteen court paintings from the Tang Dynasty. (a) Court Ladies Preparing Newly Woven Silk; (b) Spring Outing of the Court Ladies;(c) Court Ladies Wearing Flowered Headdresses; (d) Consort Yang Mounting a Horse; (e) Court Ladies in Embroidery; (f) Court Ladies Wielding Fans; (g) Play the Guqin and Savor the Tea; (h) Tuning Music to Nurture the Infant; (i) Emperor Minghuang and Consort Yang Playing Weiqi;(j) Admonishing in Chains; (k) Emperor Taizong Receiving the Tibetan Envoy; (l) Palace Ladies and Noble Children at Play; (m) Palace Ladies at Leisure; (n) Emperor Minghuang’s Journey to Su; (o) Palace Ladies Playing Double-Sixes.

Table 1.

The detailed information of the court paintings from the Tang Dynasty.

Code	Painting title	Collection location	Online digital resources
a.	Court Ladies Preparing Newly Woven Silk	Museum of Fine Arts, Boston	https://www.jiguzuo.com/guohua/zhang-xuan-dao-lian-tu.html
b.	Spring Outing of the Court Ladies	Liaoning Provincial Museum	https://www.shuge.org/meet/topic/49639/
c.	Court Ladies Wearing Flowered Headdresses	Liaoning Provincial Museum	https://www.lnmuseum.com.cn/mobile/#/about/detail?pageType=2&id=441742001591236039513557728532
d.	Consort Yang Mounting a Horse	Freer Gallery of Art	http://www.guohuajia.net/app/11-view-1844.shtml
e.	Court Ladies in Embroidery	Freer Gallery of Art	https://asia.si.edu/
f.	Court Ladies Wielding Fans	The Palace Museum, Beijing	https://www.jiguzuo.com/guohua/zhou-fang-hui-shan-shi-nv-tu-quan-juan.html
g.	Play the Guqin and Savor the Tea	Nelson-Atkins Museum of Art	https://www.nelson-atkins.org/art/collection/object/court-ladies-playing-and-drinking-tea-2116
h.	Tuning Music to Nurture the Infant	National Palace Museum, Taipei	https://digitalarchive.npm.gov.tw
i.	Emperor Minghuang and Consort Yang Playing Weiqi	Freer Gallery of Art	https://www.yfshuhua.com/22689.html
j.	Admonishing in Chains	Freer Gallery of Art	https://asia.si.edu/
k.	Emperor Taizong Receiving the Tibetan Envoy	The Palace Museum, Beijing	https://old.shuge.org/ebook/bu-nian-tu-juan/
l.	Palace Ladies and Noble Children at Play	National Palace Museum, Taipei	https://digitalarchive.npm.gov.tw
m.	Palace Ladies at Leisure	National Palace Museum, Taipei	https://www.jiguzuo.com/guohua/tang-ren-gong-yue-tu-8281-5801.html
n.	Emperor Minghuang’s Journey to Su	National Palace Museum, Taipei	https://www.360doc.cn/article/40044036_1077865827.html
o.	Palace ladies Playing Double-sixes	National Palace Museum, Taipei	https://www.jiguzuo.com/guohua/zhou-fang-nei-ren-shuang-lu-tu.html

From these paintings, 75 female figures and their attires were selected as the sample for this study. Drawing on sociology and theories of dress culture, theses attires were categorized into 6 groups based on social status, occupation, and occasions, including garden tour attires (9 samples), spring outing attires (9 samples), courtyard casual attires (15 samples), childcare attires (9 samples), weaving and dyeing work attires (15 samples), and maidservant attires (18 samples), as shown in Figure 2.

Figure 2.

The rule of clothing classification.

Experimental Equipment and Display Parameter Settings

There are inherent differences between display various devices of different brands (such as Apple, Lenovo, Huawei, etc.) in terms of RGB architecture, hardware-level color calibration mechanisms, and color management solutions. These differences may lead to systematic deviations in the extraction of RGB colors from target images, thereby affecting the accuracy of color restoration for the costumes in ancient paintings.

Specifically, Apple computers mostly use IPS (in-plane switching) panels, which offer excellent color uniformity and viewing angle stability. Some models of Huawei devices are equipped with either IPS or VA (vertical alignment) panels. Among these, VA panels tend to experience color drift when viewed from large viewing angles, causing the color parameters of the target extraction area to deviate from their true values. In terms of color gamut characteristics, Apple computers default to the P3 wide color gamut standard, which offers a wider color space coverage and more vibrant color reproduction for pigments like green and vermilion in ancient paintings. In contrast, some Huawei models primarily use the sRGB color gamut as their display standard. Under default display parameters, these models tend to present warm tones and may encounter issues with excessively high settings when extracting values for traditional clothing colors like ochre yellow and aged yellow.

Experimental Equipment

In this study, a Lenovo YOGA 14s Pro computer was chosen as the device for reading and display image. This device is equipped with a PureSight ultra-high-definition display panel. The main parameters are as follows: the resolution is 3072 × 1920 pixels (3K level), the color gamut is 100% P3 wide color gamut, the peak brightness reaches 500 nits, the color accuracy (ΔE) is less than 1, and the gamma value is close to 2.2.

The technical parameters of this display panel provide comprehensive support for the accuracy of image color extraction. Firstly, the screen is a wide color gamut of 100% P3. This color gamut fully captures the color spectrum of traditional mineral pigments and plant pigments, avoiding color clipping or approximate substitution. This restoration of subtle color transition layers in the paintings ensures that the RGB values closely match the original color spectrum characteristics. Secondly, the ΔE is less than 1. The hardware-level calibration standard ensures that the deviation between the screen-displayed colors and the standard colors is within a range imperceptible to the human eye. This enables the precise reproduction of the subtle color differences in ancient paintings and avoids extraction errors caused by color deviations in the equipment. Thirdly, the peak brightness reaches 500 nits. This brightness parameter enhances the signal-to-noise ratio of the color signal, highlights color details and grayscale levels, and prevents the loss of color information in dark areas. Fourthly, the gamma value is approximately 2.2, which closely aligns with the brightness perception characteristics of the human visual system.

Display Parameter Settings

To further ensure the authenticity and reliability of image color extraction, a set of strict standardization settings for the computer display parameters has been implemented. The specific procedure is as follows:

Step 1: The operational environment is maintained under soft lighting conditions to avoid direct light and strong reflections that could interfere with the displayed colors.

Step 2: Start the device and let it run continuously for 30 min. Once the screen’s color performance has stabilized, proceed with the color calibration operation to avoid color drift.

Step 3: Disable the built-in special display modes of the device, including eye-protection mode, color enhancement mode, and HDR (high dynamic range) display mode, to eliminate the interference from the system’s color modulation.

Step 4: Set the screen brightness and contrast parameters to 50% of their standard value. This ensures that the device consistently displays images based on a reliable hardware performance benchmark, thereby enhancing the repeatability of the color extractions.

Color Feature Analysis

The study employed the K-means clustering algorithm to extract the color features of individual clothing items and categories. These color features were further analyzed based on the COLORO system, and then the rules of color matching of various types of clothing in terms of hue, lightness, and chroma were summarized. The specific procedure for color feature analysis is illustrated in Figure 3.

Figure 3.

The specific procedure for color feature analysis.

Color Restoration

The Tang Dynasty lasted for a 1000 years. These paintings from the Tang Dynasty generally exhibit color fading, which interferes with the precise analysis of clothing color characteristics. Therefore, before conducting systematic color analysis, it is essential to perform image color restoration as a preliminary step. Currently, color restoration techniques are mainly divided into two categories: traditional methods and intelligent methods. The former relies on the expertise and craftsmanship of art restorers, using image restoration software to restore colors⁹; the latter, based on reference images, applies convolutional neural network technology to design specific network structures, computational functions, constraints, and iterative algorithms to achieve automated color restoration.¹⁰ However, both methods have limitations. Traditional restoration effectiveness heavily depends on the experience and skill level of the restorer, whereas the accuracy of intelligent methods is constrained by the choice of reference images. The degree of color-matching between the reference image and the restoration image directly determines the final restoration result.^11–13 Given the difficulty of obtaining reference images that are highly compatible with the paintings from the Tang Dynasty, the study employed the traditional method for color restoration.

Prior to the color restoration of the images, the causes of the fading and yellowing in paintings from the Tang Dynasty were analyzed. Firstly, the pigments used in Tang Dynasty paintings were predominantly mineral pigments, which resulted in a more pronounced graininess in the background colors, such as ultramarine and ochre, making them prone to slight fading.¹⁴ Secondly, the primary materials used for painting during the Tang Dynasty were silk and paper. These materials have a property that allows light to pass through, which can influence the colors of the clothing to be influenced by the background color, resulting in variations. For instance, a yellowish background might cause adjacent white clothing to take on a warm yellow hue, while a greenish-blue background could give light-colored clothing a bluish-green undertone.¹⁵ Thirdly, the texture and aging of the materials used in Tang Dynasty paintings, such as silk and paper, can create a triple reflection involving the background color and the color of the clothing. For instance, the aged yellow silk background would create a uniform yellow filter when extracting the colors from the clothing. Additionally, the reflective properties of mineral pigments, such as the luster of cobalt blue, can superimpose on the background, interfering with the accurate representation of the clothing’s true colors.¹⁴

In light of the reasons for the fading and yellowing of image colors, a set of image color restoration methods specifically designed for paintings on silk or paper, featuring background colors such as green, ochre yellow, and pure white, was proposed for this study according to the previous restoration methods.^9,14,16 These methods aim to address the issues of aging and yellowing in Tang Dynasty paintings and are intended to standardize the operational procedures for image restoration personnel, thereby ensuring the accurate extraction of the color characteristics of women’s clothing in Tang Dynasty images during the subsequent process. A skilled worker with over 10 years of experience in ancient image color restoration was recruited to undertake this important task. The specific operation is as follows:

Step 1: Remove image noise. Import the image into Adobe Photoshop CS6. Use the “Filter” > “Noise” > “Reduce Noise” function with an adjustable intensity setting in the range of 20 to 30. This will eliminate the noise caused by digital compression while preserving the details of the image.

Step 2: Image color restoration. First, determine the background type of the painting, which may include greenish-blue (as seen in the painting of Emperor Minghuang’s Journey to Su), ochre-yellow (as in the painting of Palace Ladies at Leisure), pure white (as depicted in the painting of Emperor Taizong Receiving the Tibetan Envoy), or aged yellow-brown. Then, adjust the image colors accordingly based on the identified background type. For the greenish-blue background, open the “Image” > “Adjustments” > “Color Balance.” In the “Shadow” section, set the red channel to +10 and the yellow channel to +5. Then, in the “Midtones” section, set the cyan channel to −8 and the blue channel to −5. For the ochre-yellow background, open the “Color Balance” tool. In the “Midtone” section, set the yellow channel to −12 and the blue channel to +8. Then, in the “Curves” function, brighten the blue channel by 5%–8%. For the pure white background, open the “Image” > “Adjustments” > “Brightness/Contrast,” set the brightness to +5 and the contrast to +3. Then, open the “Image” > “Adjustments” > “Hue/Saturation” and adjust the saturation to −2 to avoid distortion of the white. For the aged yellow-brown type, open the “Camera Raw Filter” > “Color Temperature.” Reduce the blue tint by 10–15. Then, navigate to “Hue” and adjust the reddish hue to +5 to neutralize the yellow tint in the image.

Step 3: Fine-tuning the image color. First, use the “Color Sampler” to select five different points from the face area of the image. Record their R, G, and B values, and calculate their average. Assume that the skin of the female face in Tang Dynasty paintings should be fair (i.e., each of the R, G, and B values should be greater than 225) and use this as the criterion for evaluation. Compare this with the calculated R, G, and B values of the image’s face. If all the R, G, and B values of the face’s image are greater than 225, then no fine-tuning of the image color is needed. Otherwise, repeat step 2 to achieve desired the judgment result.

Image Cropping

After completing the color restoration of the clothing, the non-clothing regions in the image were stripped away using the “Lasso” and “Crop” tools in Adobe Photoshop CS6. Subsequently, each clothing image was compressed to a resolution of 500 pixels by 250 pixels. Finally, the processed clothing images were imported into commericial mathematical software, MATLAB 2021, developed by MathWorks. The color data of R, G, and B channels from each pixel in the images was extracted in bulk for subsequent color analysis.

Color Feature Extraction

In the field of color feature extraction, existing studies can generally be categorized into two methods: manual extraction and intelligent extraction. During the early stages, designers primarily used manual methods, employing the built-in “eyedropper” tool in design software to extract colors from garments. However, these methods exhibit significant limitations. On one hand, they heavily rely on the subjective experience of the designer; on the other hand, the color features extracted using these methods tend to be relatively simplistic.

As the advancement of digital imaging technology, methods for color feature extraction have evolved to become more diverse. Representative techniques now include the octree algorithm, various clustering algorithms, and histogram-based methods.^17,18 Simultaneously, the color features have become significantly more enriched. Beyond traditional color values, new indicators such as color area ratio, color space distance, and bivariate color co-occurrence rates have been introduced.^19,20 The incorporation of these indicators of clothing color provides more scientifically grounded analytical dimensions for understanding design patterns in garment color-matching, effectively overcoming the limitations of conventional methods.

Taking into account the characteristics of various intelligent algorithms, this study adopted the widely used K-Means clustering algorithm to extract clothing color features. As a typical unsupervised machine learning algorithm, the K-Means clustering algorithm can efficiently identify and aggregate color units with similar spectral characteristics in clothing images by minimizing the within-cluster sum of squares criterion. Based on this algorithm, color features were extracted for each garment and each category of garments, respectively.

Color Features of a Single Piece of Clothing

The color features of a single piece of clothing include the hue (H), value (V), and chroma (C) of each color, as well as the area ratio for each color (i.e., color area ratio). The specific extraction steps are as follows:

First, the color values of the three channels (R, G, and B) from the clothing image were read using MATLAB 2021.

Next, the K-Means function was applied to perform a clustering analysis on the colors present in the clothing image. The number of clusters was set to three, representing a clothing item composed of three colors: one dominant color, one secondary color, and one accent colors.

Finally, the R, G, and B values and corresponding pixel counts were calculated for each color cluster. These R, G, and B values were then converted to H, V, and C values based on the COLORO system.²¹ The area ratio of each clustered color was computed according to the number of pixels in each cluster, using the following formula:

S_{i} = \frac{n_{i}}{N}

(1)

where S_i represents the color area ratio of the ith clustered color; n_i represents the number of pixels corresponding to the ith clustered color; N represents the total number of pixels across all clustered colors.

Color Features of a Single Clothing Category

The color features of a single clothing category include the H, C, and V of each color, color area ratio, color space distance, and bivariate color co-occurrence rate. The specific extraction steps are as follows:

First, all R, G, and B data from images of a specific clothing category were read using MATLAB 2021, and K-Means clustering was performed a second time. Since the color diversity within a clothing category is usually greater than that of a single piece of clothing, and drawing on clustering practices from existing literature for representative clothing types, the number of clusters for each category is set to eight, representing two dominant colors, two secondary colors, and four accent colors. The R, G, and B values, along with the corresponding pixel counts for each cluster color, were computed.

Next, the color space distance between any two clustered colors was calculated and then normalized, as described by the following formula:

D_{i, j} = \sqrt{\frac{{(\frac{R_{i} - R_{j}}{255})}^{2} + {(\frac{G_{i} - G_{j}}{255})}^{2} + {(\frac{B_{i} - B_{j}}{255})}^{2}}{3}}

(2)

where D_i,j represents the normalized Euclidean distance between the ith and jth clustered colors; R_i, G_i, and B_i represent the R, G, and B values of the ith clustered color, while R_j, G_j, and B_j represent corresponding values for the jth clustered color. Based on the pixel count of each clustered color, its respective area ratios were calculated using formula (1).

Finally, the bivariate color co-occurrence rate was computed, defined as the ratio of the frequency of two clustered colors simultaneously appearing in a single garment to the total number of garments in the specified category.²²

C_{i, j} = \frac{1}{M} \sum_{m = 1}^{M} V_{〈 Δ E_{i, m}, Δ E_{j, m} 〉}

(3)

where C_i,j represents the frequency with which the ith and jth clustered colors co-occur within the same garment across the entire category; M represents the total number of clothing samples within this specific category; $V_{〈 Δ E_{i, m}, Δ E_{j, m} 〉}$ represents instances where the ith and jth clustered colors are simultaneously matched in the m-th garment, with a color difference below a given threshold. Specifically, when both $Δ E_{i, m}$ and $Δ E_{j, m}$ are less than $Δ E$ , the ith and jth clustered colors are considered a color pair and are counted once; otherwise, the count is zero.

Analysis Based on COLORO System

COLORO is a nationally recognized color application and development management system developed by the China Textile Information Center.^23,24 The system defines color in three dimensions: hue, value, and chroma. Hue is based on a 360° cycle and is evenly divided into 10 distinct intervals: red, red-yellow, yellow, yellow-green, green, blue-green, blue, blue-purple, purple, and purple-red. Value is determined by the principle of light reflectance, with “ideal black” and “ideal white” representing the two extreme values of 0 and 100, respectively. It is categorized into three ranges: the low lightness (≤40), the medium lightness (40–65), and the high lightness (>65). Chroma reflects the degree of color saturation, ranging from 0 to 40, and is divided into low chroma (<8), medium chroma (8–22), and high chroma (≥22).

The study employed the visualization methods for multidimensional data to further interpret the color features of various types of garments. First, all clustered colors of individual garments from each garment type were input into the COLORO system to visualize the distribution of hues and the value-chroma distribution. Next, the frequency of clustered colors in terms of hue, value, and chroma was statistically analyzed, along with the frequency of primary colors across different garment types. Finally, by synthesizing the results of the color clustering analysis with the COLORO system visualization, the study summarized the color combination patterns and application rules specific to each garment category.

Figure 9(e) and ( f) illustrate that maidservant attire can generally be categorized into two color matching paradigms. The first paradigm is the combination of somber tones, represented by F1–F5, F7, F10, F11, and F16. These garments typically feature dark brown and dark gray as dominant hues, paired with similar or grayish-white tones, forming simple color combinations. The second paradigm is the combination of bright tones, exemplified by F8, F9, F13–F15, F17, and F18. These typically feature light yellow and pink as dominant hues. Specifically, when light yellow is the primary color, it is often paired with bright red; when pink is dominant, it is commonly matched with similar hues or neutral colors. The second paradigm displays more varied color combinations than the first.

Discussions

Clothing reflects the emotional value and cultural connotation of the wearer. It is closely related to the wearer’s social status, the occasion for wearing the attire, and their occupation. Based on the characteristics of the color matching in female attire at the Tang Dynasty court, this study further explores how social status, occasions, and occupations influence these color matching characteristics.

Effect of Social Status

The wearers of garden tour attire, spring outing attire, and courtyard casual attire in the Tang Dynasty belonged to the royal and aristocratic classes, thus representing the upper echelons of dress. The color schemes of their clothing reflected the aesthetic preferences of the social aristocracy. In contrast, maidservants, who wore servant attire, represented the lower strata of society and fell within the category of plebeian dress. Their clothing colors revealed the color preferences of the lower class. By comparing the color features of the clothing worn by noblewomen (garden tour attire, spring outing attire, and courtyard casual attire) with those of maidservants (maidservant attire), this study aims to investigate the mechanisms by which social status influenced the color schemes in female attire during the Tang Dynasty.

In terms of hue, Figures 4(d), 5(d), 6(d), and 9(d) illustrate the frequency of red usage in clothing. Meanwhile, Figures 4(c), 5(c), 6(c), and 9(c) provide the proportion of garments with red being the dominant color, calculated as the ratio of red-dominant samples to the total number of samples in each clothing category. As illustrated in Figure 10(a) and (b), both the ratios of red usage and red as the primary color are significantly higher in aristocratic attire than in maidservant attire. These findings are corroborated by the comparisons of different color area ratios shown in Figures 4(f), 5(f), 6(f), and 9(f). In the aristocratic attire, the top three colors by area ratio (AS1–AS3, BS1–BS3, and CS1–CS3) all display varying degrees of red. In contrast, the top 3 colors in the maidservant attire (FS1–FS3) do not include red.

Figure 10.

The color contrast between the attire of aristocratic class and that of lower-class individuals. (a) The frequency of red usage; (b) the ratio of red as the primary color.

In terms of value and chroma, comparisons of Figures 4(e), 5(e), 6(e), and 9(e) reveal that maidservant attire more prominently features dark gray tones with low value and low chroma (4%), whereas the ratio of this particular tone is zero in aristocratic attire.

This study further analyzes the underlying causes of these color differences across social status by referencing historical records from the Tang Dynasty. Red was considered a symbol of wealth and beauty, and as such, it was highly esteemed among women. For example, Bai Juyi wrote in A Poem for the Courtesan of Attendant Lu:“Fragrant saffron sweat soaks the singing scarf, and pomegranate-dyed skirts swirl in dance.”²⁵ Similarly, Zhang Wei wrote in To the Lady of Envoy Zhao:“Rouged cheeks and painted brows shine through mists like Chu clouds, and atop horseback, her pomegranate skirt blooms like peach blossoms.”²⁵ These literary examples demonstrate that people in the Tang dynasty often used the vibrant color pomegranate red to signify the magnificence of women’s attire. As a result, red emerged as a highly favored color choice among the garments of noblewomen during the Tang Dynasty.

By contrast, maidservants, as members of the lower class, faced restrictions in their color choices due to the hierarchical structure of Tang society. Although there are no explicit historical prescriptions regarding the color norms of servant attire, inferences can be drawn from the color systems associated with other social groups. For instance, the New Book of Tang: Treatise on Attire and Carriages stipulated color schemes for officials’ dress when paying homage to the emperor: vermilion for ranks 1–5, green for ranks 6–7, and dark blue for ranks 8–9.²⁶ For the daily attire of officials, purple was designated for ranks 1–3, vermilion for ranks 4–5, green for ranks 6–7, and dark blue for ranks 8–9. For the riding costumes of officials, purple was designated for ranks 1–3, crimson for ranks 4–5, green for ranks 6–7, and turquoise for ranks 8–9. Additionally, Du You’s Tong Dian recorded the color grades for female court officials, noting that low-ranking attendants typically wore garments in dark brown, while excluding colors such as red, purple, green, and navy blue.²⁷ Based on this, it can be inferred that maidservants, given their lower society status, likely wore dark blue or brown tones instead of red, adopting overall somber color schemes that are consistent with the color clustering results of this study.

Moreover, it is noteworthy that not all maidservant attire was characterized by a uniform dimness; some garments featured bright and vivid tones, as seen in samples F8, F9, F15, and F18. This study suggests that these exceptions are closely related to the social status of the master they served. Garments F8 and F9 originate from the painting “The Emperor’s Carriage Procession” (Figure 1(k)), which depicts maidservants attending to Emperor Taizong. Their clothing boldly pairs red and pale yellow, creating a vibrant palette that not only complements the formality of the imperial court setting but also symbolically accentuates royal majesty. Garment F15, depicted in the painting “Court Ladies with Waving Fans” (Figure 1(f)), belongs to the attendants serving an imperial consort. Its red-toned color scheme, layered with varying degrees of brightness, aligns with the dressing style of high-ranking concubines, visually reinforcing both the hierarchical relationships and the nuanced color harmony within the court dress system. Therefore, the diversity in the colors of maidservant attire ultimately illustrates the principle of “clothing color according to the master” within the overarching framework of social hierarchy.

Effect of Occasion

In addition to considering social status, the color coordination of court noble attire during the Tang Dynasty also showed notable adaptability to various dressing scenarios. Therefore, the study explored the differences in the color matching features of court noble attire across four distinct dressing scenarios: touring the garden and enjoying the flowers, riding horses for leisure, drinking tea and playing chess, and taking care of children.

In the scene of touring the garden and enjoying the flowers, the garden tour attire primarily featured various shades of red, showcasing different levels of value and chroma, as shown in Figure 4. Through the skillful pairing of dark and light hues, the outfit complemented the environment filled with blooming flowers, highlighting the wearer’s noble identity. In outdoor activities like horseback riding, the spring outing attire cleverly combined red with light cyan, creating a striking color contrast, as shown in Figure 5. This combination not only echoed the vitality of spring but also highlighted the noblewoman’s graceful and dynamic appearance through the lively contrast of colors. In an atmosphere of leisure and entertainment, characterized by activities such as drinking tea and playing chess, red, with medium value and chroma, still dominated the primary color usage. However, soft tones like grayish white, light yellow, and light green also frequently appeared (Figure 6). These tones harmonized with the scene’s atmosphere, conveying a sense of calm and refined elegance in daily life. Childcare attire, designed for daily tasks such as breastfeeding and caring for children in the imperial harem, differed from the clothing worn for social occasions, such as touring the garden, enjoying the flowers, riding horses for leisure, drinking tea, and playing chess. The psychological needs of the wearer have shifted from a focus on external beauty to an emphasis on practical functionality. Thus, the color scheme no longer used red as the primary color for clothing. As shown in Figure 7, the hues of childcare attire became more subdued, including shades such as grayish white and light yellow, characterized by higher value and lower chroma. This color strategy reduced the decorative aspect of the clothing and instead created a warm and approachable atmosphere for childcare. It demonstrated the functional transformation and humanized design of court attire in specific contexts.

In conclusion, the color coordination of attire can vary significantly depending on the occasion, even among individuals belonging to the same noble class. The variation reflects the design wisdom and cultural depth of court fashion during the Tang Dynasty, embodying the principle of “clothing color according to the occasion.”

Effect of Occupation

This study investigates how the occupations of wearers influence clothing color schemes by examining the color features of the weaving and dyeing work attire.

According to Six Codes of Tang Dynasty, the weaving and dyeing bureau operated six dyeing workshops, each responsible for producing one of the following dyes: red, blue, yellow, white, purple, and black.²⁸ Likewise, the rich color diversity observed in Figure 8 reflects the advanced dyeing techniques used in Tang Dynasty garments. However, regarding occupational categories, the color schemes of weaving and dyeing work attire are diverse, but there is little evidence of significant differences when compared to other types of clothing. Moreover, Figure 1(a) depicts groups of women engaged in three different occupations: beating, sewing, and ironing. The drying group includes six women, the sewing group two women, and the beating group includes four women. The color schemes of their garments correspond to E1–E6 samples, E7–E8 samples, and E9–E12 samples in Figure 8(c), respectively. Clearly, there are no marked distinctions in clothing color schemes among the different occupational groups.

Therefore, it can be concluded that during the Tang Dynasty, specific occupational categories, such as weaving and dyeing, had not yet developed standardized paradigms for clothing color coordination. Instead, clothing color schemes were still primarily determined by the personal preferences of the wearer.

Design and Evaluation

Design of Clothing Color Scheme

This study is grounded in the ChatGPT-4o model and centers on six distinct Hanfu categories, namely garden tour attire, spring outing attire, courtyard casual attire, childcare attire, weaving and dyeing work attire, and maidservant attire, conducting an in-depth design analysis of the color scheme in the style of the Tang Dynasty. Compared to the model of ChatGPT, ChatGPT-4o has a natively integrated text-to-image (T2I) function. This function was officially launched on March 26, 2025, and does not rely on third-party generation models such as DALL E3. It enables a native integration mode for T2I technology, significantly enhancing the coherence and operational convenience of design creation. Its technical characteristics are highly compatible with the design requirements of Tang-style Hanfu.^29,30

The intelligent generation design process of Tang-style Hanfu consists of four steps, with specific implementation path outlined as follows:

Step 1: Automatically generate textual instruction for the design draft of Tang-style Hanfu. The textual instruction consists of three major modules: historical form characteristics, scene element representation, and technical parameter specifications. Taking the garden tour attire as an example, the historical form characteristics module requires a clearly definition of the typical patterns of Tang Dynasty clothing, such as the style of tunics and skirts, the decoration of shawls, and the design of low-cut collars; the scene element representation module requires defining the design context of the scene, which includes the setting of the garden, as well as the main image of a young woman (for example, holding a round fan in her left hand and a peony flower in her right hand); the technical parameter specifications module clearly outlines the output format requirements, including a transparent background and the presentation of black lines, among other specifications.

Step 2: Construction and quantification of the color template for Tang-style Hanfu. Firstly, based on the research results regarding the color features of female clothing in the Tang Dynasty court, a three-color structure (primary color, secondary color, and accent color) was established for the color template. Each clothing style corresponds to a set of exclusive color templates. By combining the color feature analysis results shown in Figures 4(c), 5(c), 6(c), 7(c), 8(c), and 9(c), the color combination schemes for various types of clothing were clarified. The primary color, secondary color, and accent color of garden tour attire correspond to A2, A3, and A8, respectively; those of spring outing attire are B2, B5, and B7; those of courtyard casual attire are C2, C3, and C11; those of childcare attire are D1, D6, and D9; those of weaving and dyeing work attire are E4, E6, and E10; and those of maidservant attire are F1, F5, and F9. Secondly, the standard color value codes for each color in the color template were obtained through text instructions. Taking the garden tour attire as an example, its primary color value is #d99391, the secondary color value is #afa88c, and the accent color value is #d8b89f.

Step 3: Perform intelligent coloring of the clothing line drawings based on the color template. The textual instructions consist of two parts: precise annotations of color values and descriptions of texture features. Taking the garden tour attire as an example, the color value precise annotation module clearly defines the primary color as #d99391, the secondary color as #afa88c, and the accent color as #d8b89f; the texture feature description module outlines visual requirements, including the presentation of a high-quality three-dimensional form, the restoration of the luster, texture, and the interplay of light and shadow on silk, as well as the retention of the features of Tang Dynasty clothing.

Ultimately, this study utilized the ChatGPT-4o model to generate 18 color schemes for Tang-style Hanfu. Specifically, there are three different color schemes for each of the six types of Hanfu styles, as shown in Figure 11.

Figure 11.

The color schemes of six types of female court attires in Tang Dynasty. (a) Garden tour attire; (b) spring outing attire; (c) courtyard casual attire; (d) childcare attire; (e) weaving and dyeing work attire; (f) maidservant attire.

Clothing Color Evaluation

The study conducted a comprehensive evaluation of color schemes in Tang-style Hanfu using three criteria: color harmony, aesthetic appeal, and historical appropriateness. Specifically, color harmony was treated as an objective metric, calculated using an empirical formula, while aesthetic appeal and historical appropriateness were assessed subjectively through a Likert scale.

Color Harmony

The color harmony reflects the overall balance of a color scheme and serves as a guideline for applying formal aesthetic principles to product color design.³¹ It includes two key factors: the order factor and the complexity factor.³² The calculation formula is as follows:

M = O / C

(4)

where M represents the color harmony; O represents the order factor, and C represents the complexity factor. The O and C can be further expressed as follows:

C = C_{m} + C_{h} + C_{v} + C_{c}

(5)

where C_m represents the numbers of colors; C_h, C_v, and C_c represent the numbers of pairs of colors with differences in hue, value, and chroma, respectively.

O = O_{a} + O_{h} + O_{v} + O_{c}

(6)

where O_a represents the area balance of pair of colors; O_h, O_v, and O_c represent the order factors of pairs of colors in hue, value, and chroma, respectively. According the empirical parameters obtained from the study by Moon and Spencer, the area balances of primary color and secondary color, primary color and accent color, and secondary color and accent color were designated as 0.5, 0.5, and 0.25, respectively.³³ Thus, in this study, the sum of area balance from the color combinations was equal to 1.25. The O_h, O_v, and O_c are determined by the difference in hue, value, and chroma between the pairs of colors.³⁴ The order factors corresponding to various difference ranges in hue, value, and chroma are shown in Table 2.

Table 2.

The order factors of color schemes under the ranges of differences in hue, value, and chroma.

Measurements	Identity	First-ambiguity	Similarity	Second-ambiguity	Contrast
Hue
Order value	1.5	0	1.1	0.65	1.7
Difference range	[0, 4]	[5, 11]	[12, 20]	[21, 45]	[46, 80]
Value
Order value	−1.3	−1	0.7	−0.2	3.7
Difference range	[0, 2]	[3, 5]	[6, 15]	[16, 25]	[26, 100]
Chroma
Order value	0.8	0	0.1	0	0.4
Difference range	[0, 3]	[4, 9]	[10, 16]	[17, 23]	[24, 40]

For color schemes composed of multiple colors, the measure of color harmony needs to be adjusted based on the theoretical framework developed by Moon and Spencer.³³ The adjusted formula is as follows:

Z_{M} = {\begin{matrix} 0 \\ M \\ 0.5 + (M - 0.5) / (2 m - 1) \end{matrix} \begin{matrix} , \\ , \\ , \end{matrix} \begin{matrix} M < 0 \\ 0 \leq M < 0.5 \\ M \geq 0.5 \end{matrix}

(7)

where Z_M represents the normalization value of M. The Z_M is calculated based on the hue, value, and chroma of each color in the templates shown in Figure 11, using formulas (4) through (7). A Z_M value greater than 0.5 indicates a higher degree of color harmony, whereas a value below this threshold suggests lower harmony.

Aesthetic Appeal and Historical Appropriateness

The aesthetic appeal reflects the visual sensory experience evoked by the color schemes of Tang-style Hanfu. Historical appropriateness refers to the extent to which the color scheme of Tang-style Hanfu align with the cultural characteristics from the Tang dynasty.

According to ISO 11056: 2021, “Sensory analysis—Methodology—Magnitude estimation method,” which stipulates that the number of evaluators with relevant experience should not be fewer than 15, a total of 30 participants were recruited for the subjective evaluation. The participants were divided into two groups: the expert group and the general group. The expert group consisted of 5 Hanfu designers and 5 researchers specializing in ancient clothing, while the general group included 20 experienced Hanfu enthusiasts, each with over 5 years of experience in renting or purchasing Hanfu. A subjective evaluation was conducted using a 5-point Likert scale. For aesthetic appeal, scores ranged from 1 to 5, corresponding to “very unattractive,”“unattractive,”“neutral,”“attractive,” and “very attractive.” For historical appropriateness, scores ranged from 1 to 5, representing “completely inconsistent,”“somewhat inconsistent,”“neutral,”“somewhat consistent,” and “very consistent,” respectively.^35,36

The experiment comprised two phases: the preparation phase and the evaluation phase. In the preparation phase, the experimenter explained the objectives, procedures, and evaluation scales to the participants in detail. The experiment commenced upon the signing of informed consent forms. During the evaluation phase, participants sat quietly in front of a computer screen, where images of garments from Figure 11 were presented sequentially, each displayed for 10 s. Participants then rated each image based on two evaluation criteria: garment aesthetic appeal and historical appropriateness, within approximately 10 s. After each rating, the next image appeared, and the process continued until all images were evaluated. The total duration of the experiment was approximately 6 min. If any participant experienced discomfort during the experiment, it was immediately terminated.

Generally speaking, the evaluation results from the expert group are more reliable than those from the general group. Therefore, the evaluation results from the two groups are weighted in a ratio of 0.6–0.4. The aesthetic appeal (B) and historical appropriateness (F) are expressed as:

B (o r F) = 0.6 \times B_{e} (o r F_{e}) + 0.4 \times B_{p} (o r F_{p})

(8)

where B_e and B_p represent the average values of aesthetic appeal from the expert group and general group, respectively; F_e and F_p represent the average values of historical appropriateness from the expert group and general group, respectively.

To enable the analysis of aesthetic appeal, historical appropriateness, and color harmony on a unified scale, the latter two indicators need to be normalized using the following formulas:

Z_{B} (o r Z_{F}) = \frac{B - B_{min}}{B_{max} - B_{min}} o r (\frac{F - F_{min}}{F_{max} - F_{min}})

(9)

Evaluation Results

Finally, the results of color harmony (M and Z_M), aesthetic appeal (Z_B), and historical appropriateness (Z_F), as measured using the color templates shown in Figure 11, are presented in Table 3. It can be observed that, except for Z_H values in the FP1 and FP2 samples, all Z_M, Z_H, and Z_F values exceed 0.5. This indicates that the color templates derived from and constructed based on Tang Dynasty costumes exhibit a relatively high level of color harmony, aesthetic appeal, and historical appropriateness. The relatively low Z_H values in FP1 and FP2 can be attributed to the social status of the maidservants, which results in their clothing being dark, dull in color, and not overly ornate. Furthermore, the Z_F in DP3 shows a relatively low value, which may render it unsuitable for childcare attire during the Tang Dynasty due to its dull color. In conclusion, the color matching method for Tang-style Hanfu proposed in this study has received positive feedback regarding its practical application.

Table 3.

The evaluation results of color harmony, aesthetic appeal, and historical appropriateness.

Results	AP1	AP2	AP3	BP1	BP2	BP3	CP1	CP2	CP3
M	1.12	1.11	1.00	0.80	0.64	0.78	0.63	1.03	0.69
Z_M	0.62	0.62	0.60	0.56	0.53	0.56	0.53	0.61	0.54
Z_H	0.85	0.83	0.63	0.77	0.79	0.69	0.72	0.75	0.65
Z_F	0.81	0.77	0.65	0.74	0.71	0.64	0.70	0.69	0.63
Results	DP1	DP2	DP3	EP1	EP2	EP3	FP1	FP2	FP3
M	0.82	0.58	0.63	0.89	0.89	0.69	1.12	1.24	0.78
Z_M	0.56	0.52	0.53	0.58	0.58	0.54	0.62	0.65	0.56
Z_H	0.67	0.74	0.62	0.74	0.70	0.66	0.47	0.48	0.53
Z_F	0.62	0.65	0.51	0.69	0.70	0.67	0.56	0.53	0.57

Conclusion

In this study, the focus is on the color-matching features of female court attire during the Tang Dynasty. Firstly, according to social status, the occasion for dressing, and the occupational roles of the wearers, the attires were classified into six categories: garden tour attire, spring outing attire, courtyard casual attire, childcare attire, weaving and dyeing work attire, and maidservant attire. Next, a K-Means clustering algorithm was applied to extract color-matching features from each category of attire. Based on the COLORO system, an in-depth analysis of color schemes was conducted in terms of hue, value, chroma, and the coordination between dominant and secondary colors. Finally, by integrating historical literature from the Tang Dynasty, the differences in color features across garment types were analyzed from three perspectives: social class, dressing occasion, and occupation, as well as the underlying sociocultural driving forces. This establishes a reciprocal and complementary relationship between the case analysis of color schemes in garments and broader sociocultural trends.

The key findings are presented as follows:

The color schemes of female court attire during the Tang Dynasty were closely tied to the wearers’ social status. Generally, noblewomen favored red tones and vibrant color combinations, while the maidservant attire was dominated by dark blue or brown hues. Moreover, the color schemes for maidservant attire adhered to the principle of “clothing color according to the master.” For example, those serving emperors, empresses, or imperial consorts wore brighter colors, whereas others wore duller hues.

The color schemes also varied significantly depending on the dressing occasion. Noblewomen had a degree of autonomy in selecting attire suitable for different settings. For instance, during flower-viewing outings, red tones were predominant; when going on spring excursions, red was used as the primary color, with accents of teal green and warm yellow to emphasize the vitality of spring. In leisurely courtyard settings, alternating red and achromatic colors conveyed an atmosphere of comfort and elegance. When caring for young children, garments in soft, warm tones contributed to a cozy and harmonious environment.

For occupations such as beating, dyeing, and embroidery, no significant differences in color schemes were observed. This may be due to limitations in the sample size of the study. More robust conclusions would require further verification through a broader range of historical data.

Additionally, the ChatGPT model was utilized to generate innovative color schemes for Tang-style Hanfu, based on distilled color templates from various types of attires. Three evaluation metrics, color harmony, artistic appeal, and historical appropriateness, were introduced to comprehensively assess the generated color schemes. Ultimately, it was demonstrated that this color-matching method for clothing could provide designers with effective color schemes. Grounded in quantitative scientific analysis, this study systematically elucidates the artistic principles behind the color schemes of court attire during the Tang Dynasty, enhances the theoretical framework of Tang-era women’s fashion, and offers methodologies that are historically traceable and digitally applicable methodologies for modern Hanfu design.

Footnotes

ORCID iD

Rongfan Jiang

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was finanically supported by the Humanities and Social Sciences Youth Fund of Ministry of Education of China (23YJC760044), the Humanities and Social Sciences Planning Fund of Ministry of Education of China (21YJA760079), the Natural Science Basic Research Project of Shaanxi Province of China (2024JC-YBMS-789), the Social Sciences Project of Shaanxi Province, China (2024J022), and Xi’an Social Science Fund (25TY41).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Shen

CW.

Study of ancient Chinese clothing and ornaments. Commerical Press, 2011.

Chen

Jiang

Zhang

, et al. Gorgeous bright and colorful: fashion of ancient Chinese women’s clothing. SDX Joint Publishing Company, 2019.

Dai

Wang

XQ.

The comment on the color in Tang Dynasty. J Hexi Univ 2005; 6: 106–108.

Fan

Research on Ruqun in Tang Dynasty and its application on fashin design. Fujian Normal University, 2010.

Zhang

YL.

Clothing color art of Dunhuang frescoes in the early Tang Dynasty. Dunhuang Res 2005; 6: 60–62.

Tian

Research of womens’ Ru and shirt in Tang Dynasty. Chinese National Academy of Arts, 2017.

CY.

Fashion of clothing in Tang Dynasty. China Social Science Press, 2009.

Wei

Chen

Meng

Evolution characteristics and aesthetic perception of costume colors of murals in Mogao Grottoes in the Tang Dynasty. J Silk 2022; 59(9): 137–144.

YM.

Research on the virtual restoration of ancient murals based on Photoshop. J Southwest Univ (Nat Sci Ed) 2015; 37(11): 190–196.

10.

Pei

S-C

Zeng

Y-C

Chang

C-H.

Virtual restoration of ancient Chinese paintings using color contrast enhancement and lacuna texture synthesis. IEEE Trans Image Process 2004; 13(3): 416–429.

11.

Yin

Zhu

XF.

Color restoration of mural image based on double constrain convolutional neural network. J Huazhong Univ Sci Technol (Nat Sci Ed) 2020; 48(6): 6–13.

12.

Zhang

Bai

Zeng

, et al. Supporting historic mural image inpainting by using coordinate attention aggregated transformations with U-Net-based discriminator. npj Herit Sci 2025; 13(1): 1–12.

13.

Guan

Cai

, et al. Progressive generative mural image restoration based on adversarial structure learning. npj Herit Sci 2025; 13(1): 1–12.

14.

Xiaosong

Jingchao

Ziru

, et al. Exploring the restoration patterns of color in Tang and Song Dynasties paintings: Luohan Laundering as an Example. J Zhejiang Univ (Humanities Soc Sci) 2025; 55(8): 63–72.

15.

Xiaojuan

Jun

Costume restoration of maid holding a box in the tomb of princess Yongtai. Basic Sci J Text Univ 2021; 34(1): 61–66.

16.

Xiangnan

Based on the color adjustment of research on method of the restoration of Chinese paintings. Tianjin University, 2009.

17.

Liu

Jia

BQ.

Review on the study of color design. Packag Eng 2021; 42(14): 92–101.

18.

Zhang

Ren

Liu

Research on the color of Chinese landscape painting and its application in fashion design. J Text Inst 2024; 2: 1–16.

19.

Liu

Sun

, et al. Color adjacent network for product color design. Comput Integr Manuf Syst 2019; 25(9): 2355–2365.

20.

Cao

, et al. Automatic exploration and transfer design of associative rules in She ethnic clothing coloration. Multim Tools Appl 2024; 84(12): 10269–10289.

21.

Jin

Zhang

Study on color depth based on different formulas in COLORO color system. J Tex Res 2021; 42(5): 90–96.

22.

Jia

Cao

, et al. The color composition analysis of She costumes based on the nexus network model. J Silk 2022; 59(11): 100–107.

23.

Color matching method based on nine color gamut tone model. Wool Text J 2024; 52(12): 117–121.

24.

Exploration and practice of COLORO in textile color management. Dyeing Finish Technol 2024; 5: 76–79.

25.

Cao

All Tang Poetry. Zhonghua Book Company, 1960.

26.

Song

YX.

New Book of Tang. Zhonghua Book Company, 1975.

27.

Tong Dian. Zhonghua Book Company, 1988.

28.

Zhang

, et al. Six codes of Tang Dynasty. Zhonghua Book Company, 1992.

29.

Tang

Product design and evaluation methods based on AI-generated content. J Graph 2024; 45(6): 1277–1289.

30.

Liu

Zhang

Jonathan Wu

. Toward AI fashion design: an attribute-GAN model for clothing match. Neurocomputing 2019; 341: 156–167.

31.

Moon

Spencer

DE.

Geometric formulation of classical color harmony. J Opt Soc Am 1944; 34: 46–64.

32.

Moon

Spencer

DE.

Aesthetic measure applied to color harmony. J Opt Soc Am 1944; 34(4): 234–243.

33.

Moon

Spencer

DE.

Area in color harmony. J Opt Soc Am 1944; 34(2): 93–104.

34.

Tsai

Chou

JR.

Automatic design support and image evaluation of two-coloured products using colour association and colour harmony scales and genetic algorithm. Comput Aided Des 2007; 39(9): 818–828.

35.

Hao

Jiang

RF.

Tang dynasty costume design characteristic factor value analysis and its application in modern academic dress. Wool Text J 2024; 52(12): 69–76.

36.

Chen

Sh.

Research on color design and evaluation method of cultural creative products based on color harmony theory. J Northwestern Polytech Univ 2020; 38(4): 766–774.