【正文】 Computer science and technology XXXXXXTutor XXXXXXXAbstract: This paper describes the image retrieval algorithm based on color features and gives procedures to achieve. First introduced the contentbased image retrieval technology, the history of the development and contentbased image retrieval technology characteristics and the main application explored some of the key technology used in the field, and on this basis. The article focuses on the image39。s color space, image feature extraction and image similarity measure, etc., and use matlab technology to achieve a simple image retrieval procedures. The article closes with an exposition based image retrieval technology research focus and future direction of development.Keywords: Image Retrieval, features of color,color histogram, similar methods ofmeasurement圖像檢索程序運(yùn)行配置說明書一、 機(jī)器配置環(huán)境：1．硬件環(huán)境配置：CPU: GHZ 以上內(nèi)存：1G以上硬盤：20GB以上2. 軟件環(huán)境配置：本程序使用的操作系統(tǒng)為Windows XP/7，開發(fā)平臺為MATLAB 二、MATLAB ：根據(jù)軟件提示安裝即可三、運(yùn)行程序（1）MATLAB （例如：C:\MATLAB7\work）；（2）第一次運(yùn)行程序時在D盤按如下路徑建立文件夾（D:\graduate design\test_pictures），并將示例圖片放入test_pictures文件夾中；（3）打開MATLAB ，點擊fileopen,；（4）按F5運(yùn)行程序，在彈出窗口中選擇test_pictures文件夾中的一個圖片作為待匹配圖片，然后程序會自動運(yùn)行并顯示結(jié)果。源代碼：%**************************************************************************% 基于顏色特征的圖像檢索%**************************************************************************close all。 clear all。clc。[FileName,PathName] = uigetfile(39。*.*39。)。%彈出對話框得到打開的路徑Image=imread(strcat(PathName,FileName))。%打開得到的路徑下的文件[M,N,O] = size(Image)。[h,s,v] = rgb2hsv(Image)。H = h。 S = s。 V = v。h = h*360。%將hsv空間非等間隔量化：% h量化成8級；% s量化成3級；% v量化成3級；for i = 1:Mifh(i,j)=20||h(i,j)amp。；H(i,j)=0。；end；ifh(i,j)=40amp。amp。；H(i,j)=1。；end；ifh(i,j)=75amp。amp。；H(i,j)=2。；end；ifh(i,j)=155amp。amp。am；H(i,j)=3。；end；ifh(i,j)=190amp。amp。amif h(i,j)=20||h(i,j)315H(i,j) = 0。endif h(i,j)=40amp。amp。h(i,j)20H(i,j) = 1。endif h(i,j)=75amp。amp。h(i,j)40H(i,j) = 2。endif h(i,j)=155amp。amp。h(i,j)75H(i,j) = 3。endif h(i,j)=190amp。amp。h(i,j)155H(i,j) = 4。endif h(i,j)=270amp。amp。h(i,j)190H(i,j) = 5。endif h(i,j)=295amp。amp。h(i,j)270H(i,j) = 6。endif h(i,j)=315amp。amp。h(i,j)295H(i,j) = 7。endendendfor i = 1:Mfor j = 1:Nif s(i,j)=amp。amp。s(i,j)0S(i,j) = 0。endif s(i,j)=amp。amp。s(i,j)S(i,j) = 1。endif s(i,j)=1amp。amp。s(i,j)S(i,j) = 2。endendendfor i = 1:Mfor j = 1:Nif v(i,j)=amp。amp。v(i,j)0V(i,j) = 0。19if v(i,j)=amp。amp。v(i,j)V(i,j) = 1。endif v(i,j)=1amp。amp。v(i,j)V(i,j) = 2。endendend%將三個顏色分量合成為一維特征向量：L = H*Qs*Qv+S*Qv+V；Qs,Qv分別是S和V的量化級數(shù), L取值范圍[0,71]%取Qs = 3。 Qv = 3for i = 1:Mfor j = 1:NL(i,j) = H(i,j)*9+S(i,j)*3+V(i,j)。endend%計算L的直方圖for i = 0:71Hist(i+1) = size(find(L==i),1)。endT0=Hist。% 循環(huán)讀入圖像A=dir(39。D:\graduate design\test_pictures\*.jpg39。)。for k=1:size(A)B=strcat(39。D:\graduate design\test_pictures\39。,A(k).name)。Image=imread(B)。[M,N,O] = size(Image)。% 計算每一幅圖像的顏色直方圖[h,s,v] = rgb2hsv(Image)。H = h。 S = s。 V = v。h = h*360。%將hsv空間非等間隔量化：% h量化成8級；% s量化成3級；% v量化成3級；for i = 1:M20if h(i,j)=20||h(i,j)315H(i,j) = 0。endif h(i,j)=40amp。amp。h(i,j)20H(i,j) = 1。endif h(i,j)=75amp。amp。h(i,j)40H(i,j) = 2。endif h(i,j)=155amp。amp。h(i,j)75H(i,j) = 3。endif h(i,j)=190amp。amp。h(i,j)155H(i,j) = 4。endif h(i,j)=270amp。amp。h(i,j)190H(i,j) = 5。endif h(i,j)=295amp。amp。h(i,j)270H(i,j) = 6。endif h(i,j)=315amp。amp。h(i,j)295H(i,j) = 7。endendendfor i = 1:Mfor j = 1:Nif s(i,j)=amp。amp。s(i,j)0S(i,j) = 0。endif s(i,j)=amp。amp。s(i,j)S(i,j) = 1。endif s(i,j)=1amp。amp。s(i,j)S(i,j) = 2。endendendfor i = 1:Mfor j = 1:Nif v(i,j)=amp。amp。v(i,j)0V(i,j) = 0。21if v(i,j)=amp。amp。v(i,j)V(i,j) = 1。endif v(i,j)=1amp。amp。v(i,j)V(i,j) = 2。endendend%將三個顏色分量合成為一維特征向量：L = H*Qs*Qv+S*Qv+v；Qs,Qv分別是S和V的量化級數(shù), L取值范圍[0,71]%取Qs = 3。 Qv = 3for i = 1:Mfor j = 1:NL(i,j) = H(i,j)*9+S(i,j)*3+V(i,j)。endend%計算L的直方圖for i = 0:71Hist(i+1) = size(find(L==i),1)。endT=Hist。% 計算圖像庫中每一幅圖像與查詢例子圖像的歐式距離diff = (T0T).^2。EulerDistance = sqrt( sum( diff(:) ) )。w(k)=EulerDistance。end%按顏色的相似性大小顯示查詢結(jié)果圖片[B,IX]=sort(w,2)。A=dir(39。D:\graduate design\test_pictures\*.jpg39。)。for n=1:10subplot(3,4,n)。I=imread(strcat(39。D:\graduate design\test_pictures\39。,A(IX(n)).name))。 imshow(I)。end22