2012年-- 服裝設(shè)計(jì)外文翻譯--認(rèn)知色彩主題設(shè)計(jì)織物設(shè)計(jì)的新方法（節(jié)選）

1、<p>　　中文5500字，3200英文單詞，16500英文字符</p><p>　　出處：Zheng D, Han Y, Baciu G, et al. Design through cognitive color theme: A new approach for fabric design[C]// IEEE, International Conference on Cognitive Infor

2、matics & Cognitive Computing. IEEE, 2012:346-355.</p><p>　　Design Through Cognitive Color Theme: A new approach for fabric design</p><p>　　Dejun Zheng, Yu Han , George Baciu, Jinlian Hu</

3、p><p>　　Abstract—Fashion designers conceive color composition as an ensemble of tones, shades and tints that often resemble an abstract or natural theme. Color and tone adjustments are among the most frequent f

4、abric design operations. Color and tone style adjustments are part of a generic process of cognition involved in the creation of new fabric designs. A challenging problem in this process is given a s et of natural images

5、 associated with a subject or a theme, the problem is to discover the underlyi</p><p>　　Keywords- Fabric design, color theme, cognitive design, weave pattern, color enhancement.</p><p>　　1. INTR

6、ODUCTION</p><p>　　Human language understanding is generally characterized as a logical-deductive process with an underlying inductive learning procedure. However, learning theories cannot clearly explain how

7、 designers use language and color cognition so well in their designs to express a theme or a topic given the inherently small amount of data available to them. This lack of learning data lead to the idea that there may b

8、e innate language structures in the brain [1]. In [2], the author argues that the cognitive</p><p>　　In the cognitive process of creating new fabric patterns, it is important for designers to convey a meanin

9、gful theme or characteristic attribute of their desired design styles in order to both communicate as well as classify design styles. Inspiration for new designs can be found in many sources, such as photography, fashion

10、 magazines, art, and color swatches. Fashion designers often evaluate great design ideas by how people generally appreciate and cherish their designs.</p><p>　　Computer aided design tools are still considere

11、d to be amongst the most important design tools in the textiles and clothing industry especially with the advent of high precision weaving and knitting. However, the open problems that are yet to receive acceptance in fa

12、shion design, share a common root, that is, the understanding of the processes of natural intelligence and the cognitive functions of the brain [4] especially during the creative mode of operation of the brain as often f

13、ound in the p</p><p>　　There are two aspects in achieving the desired fabric designs: (1) choosing colors, and (2) presenting the design on a mood board. These aspects are difficult but fundamentally crucial

14、 tasks for both amateur and professional fashion designers.</p><p>　　Specifically, fashion designers rely on the effective choice of colors in the form of semantic themes of design styles. These are implicit

15、 in the color composition of fabrics. Psychological studies confirm the strong associations between colors and semantic themes. In fashion design, a color theme is used to describe which colors designers use and how colo

16、rs are combined and applied in the fabric patterns. A color theme is typically defined as a template of colors and an associated annotated des</p><p>　　A mood board is typically used to help establish a styl

17、e, a theme, an expression, an environment, the general atmosphere, or feeling. All of these attributes are qualitative in nature. A mood board often consists of objects found from pages torn from fashion magazines or dig

18、ital natural imagery that helps understand the inspiration of the design contents. The observers collect these inspirational tidbits from the mood board to help establish a general look and feel direction finalizing the

19、desig</p><p>　　Fashion designers choose color themes and present the design inspiration in a mood board for three major reasons. First, it is generally believed that certain color combinations are harmonious

20、 and pleasing, while others are not.</p><p>　　In the past two centuries, many theories of color compatibility have been proposed to describe and explain these phenomena. Hue templates, color harmony, and rat

21、ed color palette have been used in several computer vision and graphics projects [7-12]. Second, research studies on color language and color cognition revealed a correlation between color language and cognitive behavior

22、 [13, 14]. Revisiting the study further suggests that focal colors were easier to communicate accurately and to remembe</p><p>　　Unfortunately, choosing a color theme and preparing a mood board in fabric des

23、ign can be a very tedious experience. Commercial software platforms, such as Photoshop, EAT Jacquard CAD, and Penelope CAD, do not explicitly support these tasks. These are mostly designed for very precise translation of

24、 the fabric structure into machine operations for weaving or knitting. A fabric designer may have to search for suitable images and carefully edit it to obtain a weave pattern for simulation on the comp</p><p&

25、gt;　　Our main contribution consists of casting the process of cognition in fabric design as an optimization problem for generating a suitable cognitive color theme design combined with fabric texture properties. We provi

26、de a new framework for fabric design that simultaneously takes into account the design of the fabric color and the texture effects in a cognitive loop by providing operations at the level of color theme assembly.</p&g

27、t;<p>　　Our second contribution in this paper is that we propose a data-driven color theme optimization approach to facilitate color themes extraction for fabric designs. We use color mood clustering [17, 18] and

28、weave pattern entropy functions to extract color themes from abstract or natural images associated with a subject or a theme, and then apply them in a fabric design to maintain the same theme in the cognition of observer

29、s. The measure of the difference between input images and fabric designs is</p><p>　　Our system can automatically modify the colors of fabric designs by color themes. Fashion designers can also use the syste

30、m to optimize their fabric designs according to the on- line color theme database, such as found on Kuler website and COLOURLovers. The system can automatically generate new fabric designs and present the designs in mood

31、 boards that rival complex commercial designs and presentation boards even by professional designers.</p><p>　　The rest of this paper magnifies the details of our fabric design approach through a cognitive c

32、olor theme model and demonstrates the effectiveness of our method with experimental results that show strong associations between the original scene images and the automatically generated color theme.</p><p>

33、;　　II. RELATED WORK</p><p>　　This section summarizes relevant studies with respect to the color compatibility and image color theme enhancement.</p><p>　　Color compatibility. Color has intrigued

34、 philosophers since the ancient Greeks. Colors seen together to produce a pleasing affective response are said to be in harmony. Perhaps the most prominent theory of color compatibility is the notion of hue templates. De

35、signers often treat hue templates as starting points, rather than strict rules to learn and apply the color compatibility in fashion design. Matsuda’s color harmony model introduced a set of 8 hue templates and 10 tone t

36、emplates that have b</p><p>　　Color theme enhancement. Much work had been conducted on image color editing and transfer. Chang et al. [20] proposed an automatic approach to stylize an image by analyzing the

37、statistical characteristics of image categories and color transfer was performed between images within the same category to enhance the original image contents. Recent research paid more attention to what extent of the c

38、olor editing and transfer should be performed and controlled according to the desired color themes. Coh</p><p>　　III.DATASETS</p><p>　　Our work employs three datasets: iTextile library, Kuler d

39、ataset and COLOURLover dataset. The iTextile library comprises more than ten thousands of commercial fabric swatches collected from top weaving companies and fashion companies in China. The library contains high quality

40、fabric designs including Dobby and Jacquard, and materials cotton, silk, wool, and polyester. The fabric swatches are digitized by using a customized imaging system that captures the fabric pattern details accurately in

41、te</p><p>　　IV. FABRIC DESIGN APPROACH </p><p>　　The fabric texture design approach is represented by:</p><p>　　where D is the fabric texture design, I is the natural image associat

42、ed with a keyword or keywords, i.e. the subject(s), T is the set of color themes in which each theme is a set of ordered colors, f is a feature extraction function, and g is a mapping function from I to T .</p>&l

43、t;p><b>　　a</b></p><p><b>　　b</b></p><p>　　Figure 1: Machine diagram for the hydrangea fabric design function. (a) Fabric design with the hydrangea color theme keeping

44、shape features of input pictures. (b) Fabric design with the same color theme keeping texture statistical features. The high resolution fabric designs of (a) and (b) are provided in the section of the experimental result

45、s.</p><p>　　The fabric design process can be represented by a diagram wherein input and output are defined as in Figure 1. Taking the hydrangea fabric design as an example, we input a set of natural pictures

46、 of hydrangea and automatically produce an associated color theme.</p><p>　　The color theme contains a set of colors, e.g. pinkish colors with gradient variations. The process uses fabric texture patterns fr

47、om a fabric dataset and applies the color theme onto the patterns to generate new fabric patterns with different color effects as the output.</p><p>　　In Figure 1, the function of hydrangea fabric design is

48、illustrated in the form of an automated machine diagram. The machine generates two different fabric designs by using the same color theme but keeps different texture characteristics of the input pictures.</p><

49、p>　　A. Learning Color Theme</p><p>　　Given a set of natural images associated with a subject or a theme, we try to learn the principal colors in the color mood space. The process of color theme learning c

50、onsists of an offline phase and a runtime phase. We start with building three databases in an offline phase: an inspiration image database, a color theme database, and a fabric design database. The inspiration database w

51、as built by randomly choosing one hundred subjects or themes and for each of them downloading tens of hundreds of</p><p>　　We then use source images with the same subject to extract the color theme in the ru

52、ntime phase. A color theme is typically associated with a template of three or five colors, and in our case, five colors for a fabric design are used. The colors are further optimized by the color themes in the color the

53、me database and applied to the fabric patterns in the fabric database. Specifically, the process of color theme extraction from the inspiration image database is conducted as follows:</p><p>　　First, we set

54、up discrete histogram bins by quantizing colors from the inspiration images in the RGB color space using eight quantization bins per color channel. The RGB- histogram thus consists of 512 entries for each image. Histogra

55、ms for all images with the same color theme were normalized and saved in a matrix HI . For each bin in the histogram we calculate the corresponding color mood vector in the three-dimension color mood space by the empiric

56、al function E :</p><p>　　where E1 E2 and E3 are the principal factors of emotions in the color mood space, for instance, they are activity, weight and heat in the color emotion space [17], respectively; F1,F

57、2 , and F3 are the empirical functions obtained from the psychological color emotion experiments; L * , a * and b * are CIELAB coordinates; h is CIELAB hue angle and C * the CIELAB chroma. Equations of (4), (5) and (6) r

58、educe the number of color emotions scales to only three color emotion factors.</p><p>　　Second, we use the scalar product cp to describe the expected value of the color emotion vectors of the pixels in an im

59、age [22]. It contains the predicted mean score for each of the color emotion factors. cp is calculated by:</p><p>　　where hi is probability distribution of the quantized RGB vectors in the image; e is the co

60、rresponding color emotion vector for color bins in RGB space. Thus, the color theme of the image in the color emotion space is represented by:</p><p>　　where Ø is the representation function of the colo

61、r theme in color emotion space; wea , wew , and weh are the weights vectors;◎is the array multiplicationoperation, e.g.,wea◎E1 is the element-by-element product of wea and E1 .</p><p>　　Third, we extract the

62、 color theme from images in the CIELAB color space by using K-means algorithm. The images satisfy:</p><p>　　where Ç ea1, Ç ea 2, Ç ew1, Ç ew2, Ç eh1, and Ç eh 2 are controll

63、ed constants; n is the number of colors in the color theme.</p><p>　　B. Color-texture Relationship</p><p>　　Fashion designers deal with the relationship of the color and texture in the process o

64、f designing fabric patterns. The process of design is considered to be a transformation process of the color and texture, being from the inspiration image to the desired fabric pattern. It is generally believed that text

65、ures of natural materials are usually highly correlated with certain color combinations. Thus, a desired fabric pattern should contain such a correlation in an abstract form. We take into accoun</p><p>　　Fig

66、ure 2.An example of the fabric design process and its three determinants: A grass fabric pattern design. (a) is a theme image of grass. (b) is a fabric pattern template to be determined. There are three determinants for

67、designing the fabric pattern with the grass theme, showing the consistency in terms of the region size R between (c) and (d), texture T between (e) and (f), and color C between (g) and (h).</p><p>　　As an e

68、xample of texture correlation and color optimization, the grass texture is more likely to be associated with the green color than the blue color and a fine texture than a coarse texture. Therefore, a desired fabric patte

69、rn with the theme of grass should contain a fine texture in green. Also, other textures of natural objects may be contained in the scene, such as the cloud and sky. These different textures may take different regions in

70、the picture. For these reasons, there are three deter</p><p>　　V. TEXTURE CORRELATION</p><p>　　We correlate the textures of the natural image and fabric design by using several image processing

71、techniques and a set of customized distance metric of textures. First, an image segmentation technique is used to obtain regions of an image. Second, the edge map is obtained from the segmented regions and corner detecti

72、on is then performed to calculate the statistical texture features for each region in the natural image. Third, the weave pattern entropy in frequency domain is used to index textur</p><p>　　認(rèn)知色彩主題設(shè)計(jì)：織物設(shè)計(jì)的新方

73、法</p><p>　　摘要—時(shí)裝設(shè)計(jì)師將色彩組合構(gòu)思為色調(diào)、深淺和色調(diào)的集合，通常類似于抽象或自然的主題。顏色和色調(diào)調(diào)整是最常見的織物設(shè)計(jì)操作之一。色彩和色調(diào)風(fēng)格的調(diào)整是創(chuàng)造新面料設(shè)計(jì)的一般認(rèn)知過程的一部分。在這個(gè)過程中，一個(gè)具有挑戰(zhàn)性的問題是給定一組與主題或主題相關(guān)聯(lián)的自然圖像，該問題是發(fā)現(xiàn)連接由觀察者或在我們的例子中為織物設(shè)計(jì)者感知的紋理和色調(diào)的潛在認(rèn)知關(guān)系。這些關(guān)系最終導(dǎo)致與原始主題相似的顏色和音調(diào)的最佳

74、組合，在原始主題中，相對于相關(guān)聯(lián)的口頭描述定義顏色模板。我們將織物設(shè)計(jì)過程表述為一個(gè)認(rèn)知過程，該過程同時(shí)考慮多種色調(diào)以在織物中形成期望的主題和紋理特征。本文將色彩組合和紋理知識融入到織物設(shè)計(jì)過程中，以自動(dòng)生成從給定的自然主題或抽象主題中學(xué)習(xí)到的色彩組合。我們證明，顏色主題關(guān)聯(lián)的使用可以自動(dòng)生成新的織物設(shè)計(jì)，這些織物設(shè)計(jì)可以與復(fù)雜的商業(yè)設(shè)計(jì)相抗衡，否則即使是經(jīng)驗(yàn)豐富的設(shè)計(jì)師也難以生成這些復(fù)雜的商業(yè)設(shè)計(jì)。初步實(shí)驗(yàn)驗(yàn)證了該方法的有效性。關(guān)鍵

75、詞:織物設(shè)計(jì)，色彩主題，認(rèn)知設(shè)計(jì)，編織圖案，色彩增強(qiáng)</p><p><b>　　一、介紹</b></p><p>　　人類對語言理解通常是一個(gè)邏輯演繹的過程，也是一個(gè)潛在的歸納學(xué)習(xí)的過程。然而，學(xué)習(xí)理論并不能很好的解釋設(shè)計(jì)人員為何可以如此巧妙的使用語言和色彩認(rèn)知來表達(dá)他們設(shè)計(jì)中的一個(gè)主題或題目，而且只是給他們提供很少的數(shù)據(jù)。這種對數(shù)據(jù)的缺乏使人產(chǎn)生這樣的想法，這有可

76、能是大腦與生俱來的語言結(jié)構(gòu)。作者認(rèn)為潛在語言理解的認(rèn)知過程是非邏輯演繹或歸納過程，至少不是類似確定文本文檔主題的能力的基本形式。我們引入一個(gè)認(rèn)知顏色信息框架（CCI），通過一個(gè)跨學(xué)科的方法來研究在織物圖案設(shè)計(jì)過程中的認(rèn)知顏色主題。</p><p>　　在創(chuàng)造新的織物圖案的認(rèn)知過程中，表達(dá)一個(gè)有意義的主題或其所需設(shè)計(jì)風(fēng)格的特征對設(shè)計(jì)師來說非常重要，這樣方便他們進(jìn)行設(shè)計(jì)風(fēng)格的分類和交流。新設(shè)計(jì)的靈感可以存在于許多方面

77、，如攝影、時(shí)尚雜志、藝術(shù)和顏色樣本。時(shí)裝設(shè)計(jì)師通常會根據(jù)人們欣賞和珍惜他們作品的程度來評價(jià)設(shè)計(jì)理念。</p><p>　　計(jì)算機(jī)輔助設(shè)計(jì)工具仍然被認(rèn)為是最重要的設(shè)計(jì)工具，在紡織和服裝行業(yè)，特別是在高精度編織的出現(xiàn)以后。然而，在時(shí)尚設(shè)計(jì)中一個(gè)開放式的問題都有一個(gè)共同的根源，即自然智力和大腦的認(rèn)知功能過程的理解，特別是在設(shè)計(jì)中經(jīng)常出現(xiàn)的大腦創(chuàng)造性的運(yùn)作模式。這項(xiàng)研究是出于更好地理解的認(rèn)知信息過程的目的，這也適用于織物

78、設(shè)計(jì)和定性信息的傳輸，如對新穎性和美感的認(rèn)知了解。</p><p>　　要完成所需的織物的設(shè)計(jì)需要考慮2個(gè)方面：（1）選擇顏色（2）將設(shè)計(jì)呈現(xiàn)在色板上。這些方面是困難的，但是無論對于業(yè)余愛好者還是專業(yè)的時(shí)尚設(shè)計(jì)師這些都是關(guān)鍵的任務(wù)。</p><p>　　具體而言，時(shí)尚設(shè)計(jì)師依靠有效的顏色選擇的。這些都是隱含在織物的顏色組成中。心理研究證實(shí)顏色和語義主題之間有著緊密的關(guān)聯(lián)性。在時(shí)裝設(shè)計(jì)中，一

79、個(gè)顏色的主題是用來描述設(shè)計(jì)師使用哪種顏色以及顏色是如何搭配并應(yīng)用在織物圖案中的。顏色主題通常被定義為一個(gè)顏色的模板和相關(guān)的注釋說明或分類。此外，設(shè)計(jì)結(jié)果的呈現(xiàn)受到色板的影響。</p><p>　　色板通常用來輔助建立一種風(fēng)格，一個(gè)主題，一個(gè)表達(dá)，一個(gè)環(huán)境，大概的氛圍或感覺。所有這些屬性本質(zhì)上都是定性的。色板往往包括從時(shí)尚雜志上撕下來的對象或數(shù)字化的自然圖像，有助于理解設(shè)計(jì)內(nèi)容的靈感。觀察員從色板收集這些鼓舞人心的

80、碎片來輔助建立一個(gè)大概的外觀。色板全部都是關(guān)于粗略的開始，并在觀察過程中逐漸細(xì)化。</p><p>　　時(shí)尚設(shè)計(jì)師選擇色彩主題，并在色板上呈現(xiàn)設(shè)計(jì)靈感有三個(gè)主要原因。首先，一般認(rèn)為，某些顏色組合是和諧的，而其他的則不是。</p><p>　　在過去的2個(gè)世紀(jì)中，已經(jīng)提出了許多關(guān)于顏色兼容的理論來描述和解釋這些現(xiàn)象。在計(jì)算機(jī)視覺和圖形項(xiàng)目中，采用了色調(diào)模板、色彩調(diào)和以及指定的調(diào)色板。二是色彩

81、語言與色彩認(rèn)知的研究，揭示了色彩語言與認(rèn)知行為的相關(guān)性。再進(jìn)一步的研究表明，局部顏色更容易溝通。第三，最近的研究表明，擁有的語言學(xué)顏色分類將促進(jìn)識別并影響知覺判斷。顏色的認(rèn)知范疇似乎是與描述他們的語言緊密聯(lián)系在一起的。</p><p>　　不幸的是，選擇一個(gè)顏色主題，并在織物設(shè)計(jì)中準(zhǔn)備一個(gè)色板是一個(gè)非常繁瑣的過程。商業(yè)軟件平臺，如Photoshop, EAT Jacquard CAD以及Penelope CAD，

82、沒有明確支持這些任務(wù)。這些都是專為非常精確的將織物結(jié)構(gòu)翻譯成機(jī)器語言。一個(gè)織物設(shè)計(jì)人員可能需要尋找合適的圖像，并仔細(xì)地編輯它，以便在計(jì)算機(jī)上的進(jìn)行模擬或者是在編織機(jī)上生產(chǎn)。此外，為了在織物紋理實(shí)現(xiàn)所需的顏色主題往往需要做大量的試驗(yàn)來證明可能性。因此，它通常是自動(dòng)生成織物顏色的主題，并輔助構(gòu)建色板。</p><p>　　我們的主要貢獻(xiàn)包括在織物設(shè)計(jì)過程中的認(rèn)識過程，其目的是結(jié)合面料的紋理特性產(chǎn)生一個(gè)合適的認(rèn)知顏色主

83、題設(shè)計(jì)。我們提供了一個(gè)新的織物設(shè)計(jì)框架，同時(shí)考慮到設(shè)計(jì)的織物顏色和紋理的影響。</p><p>　　在本文中，我們的第二個(gè)貢獻(xiàn)是提出了一個(gè)數(shù)據(jù)驅(qū)動(dòng)的顏色主題的優(yōu)化方法，以方便顏色主題提取的面料設(shè)計(jì)。我們使用顏色的組合并利用編織圖案的熵函數(shù)從與主題相關(guān)的抽象或自然的圖像中提取顏色主題，然后將它們應(yīng)用于織物設(shè)計(jì)，以保持相同的主題。利用顏色度量方法對輸入圖像和織物設(shè)計(jì)的差異性進(jìn)行量化。輸出的顏色主題是對輸入圖像的修改版

84、本，這是通過將重心從顏色的色彩空間轉(zhuǎn)換到對主題的語義描述來完成的。</p><p>　　我們的系統(tǒng)可以自動(dòng)修改顏色面料設(shè)計(jì)的顏色主題。時(shí)裝設(shè)計(jì)師也可以根據(jù)線的顏色主題數(shù)據(jù)庫使用該系統(tǒng)來優(yōu)化他們的織物設(shè)計(jì)，就像在Kuler網(wǎng)站上發(fā)現(xiàn)一樣。該系統(tǒng)可以自動(dòng)生成新的織物設(shè)計(jì)，并提出了設(shè)計(jì)的色板。</p><p>　　本文的其余部分通過認(rèn)知的顏色主題模型和演示實(shí)驗(yàn)結(jié)果演示我們的織物設(shè)計(jì)方法的細(xì)節(jié)，表

85、明原始場景圖像和自動(dòng)生成的顏色主題之間的緊密關(guān)聯(lián)性。</p><p><b>　　二、相關(guān)工作</b></p><p>　　本節(jié)總結(jié)了關(guān)于顏色兼容性和圖像顏色主題增強(qiáng)的相關(guān)研究。</p><p>　　色彩兼容性。從古希臘人開始，色彩就引起了哲學(xué)家們的興趣。放在一起看的顏色產(chǎn)生令人愉悅的情感反應(yīng)我們認(rèn)為二者是和諧的。也許最突出的顏色兼容性理論是色

86、調(diào)模板的概念。設(shè)計(jì)師通常把色調(diào)模板視為出發(fā)點(diǎn)，而不是在時(shí)尚設(shè)計(jì)中的色彩相容性方面運(yùn)用嚴(yán)格的規(guī)則。Matsuda的顏色和諧模型引入一套8色調(diào)和10色調(diào)的模板，模板被應(yīng)用在計(jì)算機(jī)視覺和圖形學(xué)的項(xiàng)目。這套模型的提出是基于長達(dá)九年的對日本女學(xué)生的時(shí)尚調(diào)查問卷而提出的，而調(diào)查問卷是以時(shí)裝公司提供的書籍的方式進(jìn)行。最近，心理學(xué)家開始參與顏色兼容性和喜好的研究來確定顏色組合的和諧性。一個(gè)大型的顏色偏好測試和顏色兼容性調(diào)查采用從數(shù)千參與者的五色主題進(jìn)行

87、。顏色的兼容性研究是根據(jù)用戶產(chǎn)生的顏色組合進(jìn)行的。</p><p>　　顏色主題增強(qiáng)。關(guān)于圖像的顏色編輯和傳輸，我們開展了大量的工作。常等人提出了一種自動(dòng)的方法，通過分析圖像的類別對圖像進(jìn)行風(fēng)格化，顏色的轉(zhuǎn)換通過同一類別的圖像增強(qiáng)原圖像內(nèi)容的方式進(jìn)行。根據(jù)所需的顏色主題，最近的研究更加注重顏色編輯和轉(zhuǎn)換應(yīng)該被控制或執(zhí)行到何種程度。運(yùn)用色彩是一件棘手的事。你必須以恰當(dāng)?shù)姆绞皆诤线m的時(shí)間以恰如其分的理由面向目標(biāo)受眾對

88、它進(jìn)行使用?？贫鞯热烁鶕?jù)統(tǒng)一規(guī)則進(jìn)行顏色調(diào)整。最近的幾項(xiàng)研究，旨在提高圖像顏色的主題，并在抽象和自然的圖像應(yīng)用表現(xiàn)出有效性。抽象圖像的基于主題的顏色增強(qiáng)作為一個(gè)五色主題的優(yōu)化問題被提出。四色的主題是固定的，第五個(gè)顏色是通過使用顏色兼容性模型遞歸搜索顏色主題數(shù)據(jù)庫找到最佳匹配的顏色主題。對抽象和自然的圖像進(jìn)行色調(diào)風(fēng)格的增強(qiáng)和顏色主題的增強(qiáng)。由于這些增強(qiáng)的技術(shù)訓(xùn)練的顏色紋理關(guān)系從圖像數(shù)據(jù)庫中，主題的增強(qiáng)的質(zhì)量依賴于完全和固定在數(shù)據(jù)庫中的圖像

89、內(nèi)容和來源。因此，顏色主題的增強(qiáng)不能獨(dú)立工作，這些限制阻礙了時(shí)裝設(shè)計(jì)師的創(chuàng)作。因此，它是想要開發(fā)一個(gè)有效的框架，以方便織物設(shè)計(jì)。</p><p><b>　　三、數(shù)據(jù)集</b></p><p>　　我們的工作采用三個(gè)數(shù)據(jù)集：紡織品數(shù)據(jù)庫、Kuler數(shù)據(jù)集和COLOURlovers數(shù)據(jù)集。紡織品數(shù)據(jù)庫包括超過從中國頂級織造和時(shí)尚公司收集來的數(shù)十萬商業(yè)布樣。該數(shù)據(jù)庫包含高

90、質(zhì)量的織物設(shè)計(jì)，其中包含Dobby 和Jacquard以及棉花，絲綢，羊毛，滌綸等材料。通過一個(gè)定制成像系統(tǒng)能夠?qū)⒖椢飿颖緮?shù)字化，無論是從顏色還是結(jié)構(gòu)方面，該系統(tǒng)能夠準(zhǔn)確捕獲織物圖案。我們希望有這樣一個(gè)系統(tǒng)，該系統(tǒng)能夠數(shù)字化的記錄織物設(shè)計(jì)的真實(shí)細(xì)節(jié)，使織物設(shè)計(jì)的顏色主題以及織物的紋理可以通過可靠的數(shù)據(jù)集來進(jìn)行研究。在主題顏色測試種，我們使用兩種在線顏色主題數(shù)據(jù)集，即Kuler數(shù)據(jù)集和COLOURLover數(shù)據(jù)集。前者包括44986個(gè)顏色

91、的主題和后者約包含383938個(gè)顏色主題。</p><p><b>　　四、織物設(shè)計(jì)方法</b></p><p>　　織物的紋理設(shè)計(jì)方法：</p><p><b>　　(1)</b></p><p><b>　　(2)</b></p><p>　　D是織

92、物紋理設(shè)計(jì)，I代表與一個(gè)或多個(gè)關(guān)鍵詞相關(guān)的自然圖像，比如主題，T是一組有序的顏色組成的顏色主題的集合，f是一個(gè)特征提取函數(shù)，g是I到T的映射函數(shù)。</p><p><b>　?。╝）</b></p><p><b>　　（b）</b></p><p>　　圖1：為繡球織物設(shè)計(jì)的功能機(jī)圖。（a）具有繡球顏色主題并保持輸入圖像

93、的形狀特征的織物設(shè)計(jì)。（b）具有相同的顏色主題并保持紋理統(tǒng)計(jì)特征的織物設(shè)計(jì)。（a）和（b）的高分辨率織物設(shè)計(jì)在實(shí)驗(yàn)結(jié)果部分進(jìn)行展示。</p><p>　　織物設(shè)計(jì)過程可以用一個(gè)圖來表示，其中輸入和輸出被定義為圖1。以繡球花織物的設(shè)計(jì)為例，我們輸入一組繡球的照片并自動(dòng)生成相關(guān)的主題顏色。</p><p>　　顏色主題包含了一系列的顏色，如具有梯度變化特征的粉紅顏色。該方法使用從織物數(shù)據(jù)集的織

94、物紋理圖案，并將顏色的主題應(yīng)用到圖案上，以產(chǎn)生不同顏色效果的新織物圖案。</p><p>　　在圖1中，繡球花織物設(shè)計(jì)的功能以一個(gè)自動(dòng)化的機(jī)器圖的形式進(jìn)行說明。該機(jī)器使用相同的顏色主題生成2個(gè)不同的織物設(shè)計(jì)，但保持不同輸入圖片的紋理特征的。</p><p><b>　　A.學(xué)習(xí)顏色主題</b></p><p>　　給定一組與主題或主題相關(guān)聯(lián)的自

95、然圖像，我們試著在顏色空間中學(xué)習(xí)主色調(diào)。顏色主題學(xué)習(xí)的過程包括一個(gè)離線階段和運(yùn)行階段。探討色彩之間的視覺刺激差異，視覺心里感受程度差異，色彩對情感與情緒的影響差異，發(fā)現(xiàn)人們更容易接受平衡色彩的搭配組合。我們從建立三個(gè)數(shù)據(jù)庫中的離線階段開始：靈感的圖像數(shù)據(jù)庫，顏色主題數(shù)據(jù)庫和織物設(shè)計(jì)數(shù)據(jù)庫。靈感數(shù)據(jù)庫是建立在隨機(jī)選擇一百個(gè)主題，并通過谷歌或閃爍搜索標(biāo)記圖像的關(guān)鍵詞，為他們每個(gè)人下載數(shù)百個(gè)彩色圖像。顏色主題數(shù)據(jù)庫包括超過400000個(gè)代表顏

96、色的主題，他們都來自現(xiàn)有的文獻(xiàn)和網(wǎng)絡(luò)社區(qū)如Adobe Kuler和colourlover?？椢飻?shù)據(jù)庫包含紡織企業(yè)收集的一萬種織物樣品的圖像。</p><p>　　然后，我們使用在運(yùn)行階段中提取的顏色主題的源圖像。一個(gè)顏色主題通常與一個(gè)具有三種或五種顏色的模板相關(guān)，我們的設(shè)計(jì)中使用五種顏色的織物設(shè)計(jì)。顏色由顏色主題數(shù)據(jù)庫中的顏色主題進(jìn)一步優(yōu)化，并應(yīng)用于織物數(shù)據(jù)庫中的織物圖案。具體而言，從靈感圖像數(shù)據(jù)庫中提取顏色主題

97、的過程如下：</p><p>　　首先，我們通過RGB顏色空間中的靈感圖像，按照量化顏色建立了離散直方圖，按照量化顏色，每個(gè)通道使用8個(gè)量化箱。每個(gè)圖像的RGB直方圖由512項(xiàng)組成。對相同的顏色主題的所有圖像的直方圖進(jìn)行了規(guī)范，并保存在一個(gè)矩陣中。對于每一個(gè)的直方圖中，我們計(jì)算出三維顏色的空間中相應(yīng)的顏色向量的經(jīng)驗(yàn)函數(shù)E：</p><p><b>　　(3)</b>&

98、lt;/p><p><b>　　(4)</b></p><p><b>　　(5)</b></p><p><b>　　(6)</b></p><p>　　E1，E2 和E3是顏色空間中的主要的影響因素，例如他們在色彩空間中代表活動(dòng)、重量和熱；相應(yīng)的，F(xiàn)1，F(xiàn)2，F(xiàn)3是從心理情感色

99、彩實(shí)驗(yàn)中得到的經(jīng)驗(yàn)函數(shù)；L * ,a *以及b *是CIELAB坐標(biāo)；H為CIELAB色度和C CIELAB色度。方程組（4），（5）和（6）減少顏色情緒量表的數(shù)量至只有三種顏色的情感因素。</p><p>　　其次，我們使用標(biāo)量積來描述圖像中像素的顏色情感向量的期望值。它包含了每個(gè)顏色的情感因素的預(yù)測平均得分為。cp的計(jì)算方法如下：</p><p>　　cp = hi · e