外文翻譯---關于顏色識別中顏色特征分析的方法

1、<p><b>　　英文原文</b></p><p>　　2011 Eighth International Conference on Fuzzy Systems and Knowledge Discovery (FSKD) </p><p>　　An Approach of Color Feature Evaluation in C

2、olor</p><p>　　Recognition</p><p>　　Qi Xiaoxuan, Ji Jianwei</p><p>　　School of Information & Electric Engineerin</p><p>　　Shenyang Agriculture University</p>

3、<p>　　Shenyang, China</p><p>　　Abstract—This paper analyzes the characteristics of five commonly used color spaces and explores their influences on color recognition respectively. Divisibility evaluati

4、on based on distance criterion is utilized to evaluate the different colorfeatures in each color space and experimental results show that HSI color space has the best divisibility performance. Keywords-color space;colo

5、rrecognition; feature evalutation; divisibility critiron </p><p>　　I. I NTRODUCTION </p><p>　　Color is the most intuitive vision feature to describe colorful images. It has been widely use

6、d in pattern recognition for the reason that color feature is almost free from the effects of scale, rotation and translation for the input images [1]. Colors in colorful images can be defined by different color space mo

7、dels, such as RGB space, CMY space, I1I2I3space, YUV space and HSI space. Among the above color spaces, RGB is the basic and the most common one and can readily be mapped into othe</p><p>　　color perc

8、eption and is too easily influenced by light. The three color components of RGB space are correlated with each other [2]. CMY space represents colors by the complementary component of RGB components.

9、 YUV space, frequently used in color TV systems, uses three channels as Y, U and V to define the pixel. Y are the brightness information, U and V are the color difference which denotes the overall color diff

10、erence instead of the difference between the th</p><p>　　of color, which enables them more suitable for occasions where light intensity changes, than RGB space.Color recognition technique has been applied

11、to many fields and has gone ahead rapidly. For instance, color recognition in product surface, license plates identification, face recognition and skin recognition [3-6]. Color recognition effects differ w

12、ith the change of color space. This paper investigates on color feature divisibility in the commonly used color spaces as RGB</p><p>　　above color spaces based on the distance criterion. It provides a

13、 theory basis for color recognition. </p><p>　　II. COLOR SPACE AND I TS T RANSFORMATION </p><p>　　It is essential to build up and select a suitable color space for obtaining a kind of valid

14、 color features to characterize colorful images. Different color spaces are utilized for different research purposes. Color space means to define color by an </p><p>　　array in three-dimension space. In the

15、processing of colorful images, color space is also named as color model or color coordinates. One color space can be converted to another by certain transforms. Below is the introduction of some color spaces

16、 and their conversions [7]. </p><p>　　A. RGB Color Space </p><p>　　Red (R), green (G), blue (B) are three primary colors ofspectrum. All colors can be generated by the sum of the threeprimar

17、y colors. In digital images, values of R, G and B rangefrom 0 to 255. A cube in three-dimension coordinate space can be used to describe the RGB color space, where red, green andblue are the three axes, shown in Fig. <

18、;/p><p>　　1.The main drawback of RGB color space as follows: </p><p>　　? It is not intuitive. It is difficult to see from the RGBvalues the cognitive attributes that the color repres

19、entsitself. </p><p>　　? It is non-uniform. The perception difference betweentwo colors in RGB space is different from the distancebetween the two colors. </p><p>　　? It is dependent on hardw

20、are devices. </p><p>　　In a word, RGB space is device-related and an incompleteintuitive color description. To overcome these problems, othercolor spaces,which are more in line with characteristics of c

21、olor vision, are adopted. RGB space can be mapped to other color spaces readily. </p><p>　　B. CMY(CMYK) Color Space </p><p>　　CMY space is a spatial structure of a rectangular Cartesian. Its

22、three primary components are cyan (C), magenta (M) and yellow (Y). Colors are obtained by subtractive colors. CMY space is widely used in non-emission display as inkjet printers. Equal amount of the three components can

23、generate the black color. But the aforementioned black color is not pure. Generally speaking, to generate true black color, the fourth component, i.e. black, is added in. This is the CMYK color space. CMY sp</

24、p><p>　　The transformations from RGB space to CMY space are as follows: </p><p>　　C. YUV and YCrCb Color Space </p><p>　　YUV space and YCbCr space both generate a luminance componen

25、t and two </p><p>　　chrominance components. In YUV space, Y is the luminance component, U and </p><p>　　V color difference. Y component is independent of the other two. Moreover, the YUV space c

26、an reduce the storage capacity required by digitalcolorful images by the characteristic of human vision. In YCbCr space, Y is the luminance component, Cb is the blue color component and Cr the red color compon

27、ent. Its advantages are obvious that color components are separated from luminance components and linear transformation can be performed from RGB space. Transformation from</p><p>　　D. H

28、SI Color Space </p><p>　　HSI space is established from the human psychologicalperception point of view. H (hue) is a color in a color corresponding to the main wavelength in chromatography. S

29、(saturation) is equivalent to the purity of color. I (intensity ) is the brightness of color and the uniform amount of feeling. HSV (hue, saturation, value) and HSB (hue, saturation, brightness) are other color spaces si

30、milar to HSI color space , and are all belong to polar coordinate space structure. Their common mer</p><p>　　HSI color space provides a suitable space with three components that is better used to

31、descript color in line with human hobbits. However, the defect of non-linear in color difference still exists, especially the color and angle in the H component [9]. </p><p>　　E. I1I2I3 Color Space &l

32、t;/p><p>　　Linear transformation from RGB space to I1I2I3space can be explained by the following equations to get three orthogonal color features:</p><p>　　From formula (7), it can be seen that val

33、ues of I1, I2and I3 component can be positive and negative. The non-correlation property of I1I2I3 space is the best in image recognition. </p><p>　　III. FEATURE EVALUATION OF COLOR SPACE </p

34、><p>　　By color spaces, the abstract, subjective visual perception can be translated into a concrete specific position, vector in three-dimensional space, which makes it possible to visualize color fea

35、tures of colorful images and devices. Color space is an important tool of color recognition. Various mixing system has its corresponding color space, and different color spaces have different properties with th

36、eir respective advantages and disadvantages. Validity of color space is the k</p><p>　　? Calculate the mean vector and covariance of the ith class samples, N is the number of total samples and Nithe num

37、ber of the ithclass samples. </p><p>　　IV. E XPERIMENTAL RESULTS AND ANALYSIS </p><p>　　When identified by human eyes, colors are divided into eleven categories as red, green, blue, yello

38、w, purple, orange, pink, brown, gray, white and black, shown in Fig. 2. </p><p>　　The evaluation algorithm is performed respectively on RGB space, CMY space, YUV space, I1I2I3 spaceand

39、 HSI space. Feature parameter and assessment indicators are shown in Table Ⅰ.</p><p>　　Seen from Table Ⅰ, the HSI space has the best performance compared to other four analyzed color spaces. </p>

40、;<p>　　V. CONCLUSION </p><p>　　It is necessary to select an effective color space for colorful image processing. This paper analyzes and compares the performance of five common color spaces base

41、d on divisibility criterion. Experimental results show that HSI color space has the best divisibility performance compared with the other four spaces. It provides a basis for color space selection in color rec

42、ognition. </p><p>　　ACKNOWLEDGMENT </p><p>　　This work was supported by a grant from Natural Science Foundation of Liaoning Province (Grant number: 20102153). </p><p&

43、gt;　　REFERENCES </p><p>　　[1] WANG Hui, LVYan, ZHANG Ka, “Research on Color Space Applicable to Wood Species Recognition”, FORESTRY MACHINERY & WOODWORKING </p><p>　　EQUIPMENT, Vol.37, pp.

44、20-22,2009. </p><p>　　[2] Palus, H. Representations of color images in different color spaces, The Color Image Processing Handbook. London: Chapman& Hall,1998. </p><p>　　[3] WANG Y

45、an-song, JIN Wei-qi, “Surface Inspection Based on Color Clustering of Mapping Chromatism”, Transactions of BeijingInstitute of Technology, Vol.30,pp.74-78,2010. </p><p>　　[4] CAO Jian qiu, WA

46、NG Hua qing and LAN Zhang li, “Skin Color D ivision Base onModified YCrCb Color Space”, JOURNAL OF CHONGQ ING JIAOTONG UN IVERSITY( NATURAL SC IENCE) , Vol. 29, pp.488-492,2010. </p><p>　　[5] W

47、ANG Feng, MAN Li-chun and XIAO Yi-jun et al., “Color Recognition of License Plates Based on Immune Data Reduction”. JOURNAL OF SICHUAN </p><p>　　UNIVERSITY (ENGINEERING SCIENCE EDITION), vol.40, pp. 164-17

48、0, 2008. </p><p>　　[6] XU Qing, SHI Yue-xiang, XIE Wen-lan and ZHANG Zheng-zhen, “Method </p><p>　　of face detection based on improved YUV color space”, Computer Engineering and Applications, V

49、ol. 44, pp.158-162,2008. </p><p>　　[7] HAN Xiaowei, Study on Key Technologies of Color Image Processing, Northeast University,2005. </p><p>　　[8] LIU Zhongwei, ZHANG Yujin, “A Compar

50、itive and Analysis Study of Ten </p><p>　　color Feature-based Image Retrieval Algorithms”,SIGNAL PROCESSING, Vol.16, pp.79-83, 2000. </p><p>　　[9] Liu Jin, Chen Gi, Yu Ruizhao, “Development of

51、 Computer Color Science”, COMPUTER ENGINEERING,2,1997. </p><p>　　[10] YANG Shuying.Pattern recognition and Intelligent Computation: Matlab Technolgoy, Beijing: Electronic Industry Press, 2008. </p>

52、<p><b>　　中文譯文 </b></p><p>　　2011第八屆模糊系統(tǒng)與知識挖掘國際學術會議（FSKD）</p><p>　　關于顏色識別中顏色特征分析的方法</p><p><b>　　紀建偉 齊曉軒</b></p><p><b>　　信電與電氣工程學院<

53、/b></p><p><b>　　沈陽農(nóng)業(yè)大學</b></p><p><b>　　沈陽，中國</b></p><p>　　摘要：分析五種常用的顏色空間的特征并研究其分別對顏色識別的影響。基于距</p><p>　　離標準的可除性鑒定被用來評定每個顏色空間內的顏色特征，實驗結果表明HSI&l

54、t;/p><p>　　顏色空間可除性最強。 </p><p>　　關鍵詞：顏色空間；顏色識別；特征分析；可除性標準 </p><p><b>　　I． 概述 </b></p><p>　　顏色是描述色彩圖像時最直觀的視覺特征。鑒于顏色特征幾乎不受范圍，旋轉及轉化對輸入圖像的干擾，顏色被廣泛地應用于圖像識別【1】。色彩圖像中

55、的顏色可由不同的顏色空間模式規(guī)定，比如RGB空間，CMY空間，I1I2I3空間，YUV空間和HSI空間，其中，RGB是最基本也是最常見的顏色空間，并且可以很容易的映入其他顏色空間中。但是，RGB空間與顏色直覺不一致，而且過于容易被光線影響。此空間的三個顏色分量是互相關聯(lián)的【2】。CMY空間通過RGB空間分量補充性分量來表現(xiàn)顏色。常被用于彩色電視系統(tǒng)的YUV空間通過Y,U,V三個波道來定義像素。Y表示的是亮度信息，U和V是色差，并決定整體

56、色差，而RGB則是靠三個分量之間的差別來影響整體色差的。HIS空間與人類對顏色的直覺相一致，其三個分量互相獨立，并且可以分別覺察到每個分量的變化，但是HSI空間中的非線性變化可能導致當飽和度低的時候出現(xiàn)大量的運算及顏色空間的異常。然而在YCbCr顏色空間中，色度成分和亮度成分是相互依賴的。除此之外，由于從YCbCr空間到RGB空間的轉換是很簡單的線性轉化，所以前者通常被應用于視頻編碼壓縮領域。YUV空間，YCbCr空間和</p&g

57、t;<p>　　II.顏色空間及其轉化</p><p>　　建立并選擇一個適合的顏色空間對獲得一種有效的顏色特征來描述色彩圖像的特征至關重要。不同的顏色空間用于不同的研究目的。顏色空間是指用三維空間的數(shù)組來定義顏色。在色彩圖像的處理過程中，顏色空間又稱顏色模型，或顏色坐標。通過某些轉化，可以將一個色彩空間轉換為另一個。下面是一些顏色空間的介紹及其轉化。 </p><p>　　

58、A. RGB顏色空間 </p><p>　　光譜的三個原色是紅（R），綠（G），藍（B），所有的顏色都能用三種基本組合起來形成。在數(shù)碼影像中，R，G，B的取值范圍0到255。在三維坐標空間中的多維數(shù)據(jù)集可用于描述 RGB 顏色空間，紅，綠，藍分別為軸線，如圖表1所示。 </p><p>　　RGB顏色空間最主要的缺點如下所列： </p><p>　　非直觀性。很難從

59、RGB數(shù)據(jù)庫中看到顏色本身所表示的認知屬性。</p><p>　　非一致性。RGB顏色空間中兩種顏色間的感知區(qū)別與其間的距離是不同的。</p><p><b>　　依賴于硬件設備 </b></p><p>　　總之，RGB空間與設備相關，是一個不完備的直觀顏色說明。為了解決這些</p><p>　　問題，采用其他更符合顏

60、色視覺特征的顏色空間，使得RGB空間可以很容易映射</p><p><b>　　到其中。 </b></p><p>　　B．CMY(CMYK)顏色空間 </p><p>　　CMY 空間是笛卡爾直角坐標系的空間結構，其三個主要組件是青色（C），品紅（M），黃色（Y）。相減混色模式以吸收三基色比例不同而形成不同的顏色。CMY空間被廣泛應用于像噴墨

61、打印機一類的非發(fā)射顯示器。理論上來說，三種等量基色可形成黑色，但是由于目前制造工藝還不能造出高純度的油墨，CMY相加的結果實際是一種暗紅色。一般而言，要想得到高純度的黑色就要加入第四個基色，也就是黑色；這就是CMYK顏色空間。CMY 空間不是很直觀且非線性，它的三個組件是紅，綠，藍三個互相補充的顏色，在CMY模型中，顏色是從白光中減去一定成分得到的。其轉化如下所示：從RGB空間到CMY空間的轉化如下所示： </p><

62、;p>　　C．YUV和YCbCr顏色空間 </p><p>　　YUV空間和YCbCr空間都含有一個亮度元素和兩個色度元素。在YUV空間中，Y表示明亮度，而U和V表示的則是色差，它的亮度信號Y和色度信號U、V是分離的，而且，YUV 空間可以減少由人類視覺特性的數(shù)字彩色圖像所需的存儲容量。在YCbCr空間中，Y是指亮度分量，Cb指藍色色度分量，而Cr指紅色色度分量；其優(yōu)點顯而易見：顏色分量與亮度元素相分離，并

63、可從RGB空間進行線性轉換。通過以下方程式可大致解釋從RGB到YUV的轉換關系： </p><p>　　D.HSI顏色空間 </p><p>　　它反映了人的視覺系統(tǒng)感知彩色的方式，以色調（H）、飽和度（S）和強度（I）三種基本特征量來感知顏色；與其相似的還有HSV顏色空間（色相h，飽和度s，色調v）和HSB顏色空間（色相h，飽和度s，亮度b），這些都屬于極坐標空間的結構，它們共同的優(yōu)點是

64、可以直觀的描述顏色，且大多可以從RGB空間進行線性轉換。HSI模型的建立基于兩個重要的事實： 一是I分量與圖像的彩色信息無關；二是H和S分量與人感受顏色的方式是緊密相聯(lián)的，其中H分量對顏色的描述能力最接近于人類的視覺，因此，它的區(qū)分力也是最強的【8】。 RGB空間到HSI空間的轉換如下方程式所示： </p><p>　　有三元素的HSI顏色空間與人類習慣相適應，可以更好的描述顏色。但色差中依然存在非線性的缺點，尤

65、其是H分量的顏色與角度。 </p><p>　　E．I1I2I3顏色空間 </p><p>　　RGB 空間到I1I2I3 空間的線性轉換如下方程式以得到三個顏色直方特征： 由公式（7）可知，I1,I2,I3 的數(shù)值可正可負，在圖像識別中，I1I2I3空間具有最佳的非相關性。 </p><p>　　III.顏色空間的特征分析 </p><p>

66、;　　通過顏色空間，抽象、主觀的視覺感知可被轉化為三維空間中的特定位置、矢量，從而有可能使彩色圖像和設備的顏色特征可視化。顏色空間是顏色識別的重要工具，各種混合系統(tǒng)有其相應的顏色空間并有不同的屬性和各自的優(yōu)缺點。顏色空間的有效性是處理彩色圖像的關鍵，可除性標準則被用來檢測不同顏色空間對顏色的分類。距離判據(jù)具有簡明清晰的概念，即，類屬中距離越小，類屬間距離越大，可除性便越大。下面的特征分析運算法則便是基于距離判據(jù)【10】。 計算均值向量和

67、i 類樣本的協(xié)方差，N 是總樣本數(shù)，Ni 是i 類樣本數(shù)。</p><p>　　計 算 總 的 均 值 向 量和協(xié)方差 。</p><p>　　構造i 類 樣 本 的 散 射 矩陣。</p><p>　　構 建 在 同一類 樣 本 的 總 散 射 矩陣。</p><p>　　構 建 不同的 類之間 樣 本 的 總 散 射 矩陣。</p&

68、gt;<p>　　定義評價指標，如下所示： </p><p>　　IV.實驗結果及分析 </p><p>　　顏色被人的眼睛分為十一類，紅色、綠色、藍色、黃色、紫色、橙色、粉紅色、棕色、灰色、白色和黑色，如圖表2。 對RGB空間,CMY空間,YUV空間,I1I2I3空間,和HSI空間分別進行評估運算，特征參數(shù)和評價指標如表格I所示。 從表格I 可知，HSI 空間與其它四個所

69、分析的顏色空間相比最具可除性。 </p><p><b>　　V.總論 </b></p><p>　　必須選擇有效的顏色空間來處理彩色圖像。本文在可除性標準的基礎上，對五個常用的顏色空間進行分析和比較，實驗結果表明HSI 顏色空間具有最強的可除性。這也為顏色識別中的顏色空間選擇提供了基礎。 </p><p><b>　　感謝 <