507452 A7 5736twf.doc/008 R7 發明說明（（） 經濟部智慧財產局員工消費合作社印製 發明領域 (請先閱讀背面之注意事項再填寫本頁) 本發明是有關於一種彩色影像資料庫建立的方法及 彩色影像擷取的方法，且特別是有關於一種通訊系統中 傳輸影像有極大幫助之彩色影像資料庫建立的方法及彩 色影像擷取的方法。 發明背景 在數位化的世界裡充滿各種格式的資料與媒體，在 儲存處理上，必須要有許多編碼擷取及查詢等各種機制 來滿足一般或即時的擷取，其中以內涵爲主的攝取 (Content_Based Retrival，CBR)爲最主要的解決方式。 各種資料或媒體的內涵包括數位文字、音效、音樂、圖 片及動畫電影等，藉此建立成資料庫，可以提升這些資 料信號儲存管理系統的效能，因爲傳統的資料庫系統通 常是以資料本身簡單的屬性來建立或搜尋，在管理之建 立、更新及查詢的效能不佳。 另外，CBR亦可以提供類似現今廣泛使用的全球 資訊網中瀏覽媒體的機制，提供搜尋引擎的搜尋索引， 有的人在全球資訊網的搜尋引擎提供網頁文件之關鍵字 的查詢，雖然在設計上簡單且快速，但是關鍵字的查詢 有下列的缺點： (1).無法處理一些不容易描述的查詢語言，例如，” 找出有太陽的風景照片”。（2).關鍵字的錯用以致嚴重影 響到攝取的結果。（3).描述的方式是一些單一關鍵字的 排列組合，而造成查詢的不精確及不完整。（4).關鍵字 3 本紙張尺度適用中國國家標準（CNS)A4規格（210 X 297公釐） 507452 A7 5736twf.doc/008 B7 五、發明說明（1) 之資料庫需要由人預先做分類，無法自動建立資料庫。 (請先閱讀背面之注意事項再填寫本頁) 各種資料或媒體若以CBR的方式建立儲存管理系 統及完成查詢，可以增加資料本身的加値性，其中以影 像圖片，更需要以CBR的方式達到有效的儲存及攝取的 功能。 在CBR對影像圖片的硏究中有許多方式可以完成 查詢，如上所示，關鍵字、影像邊緣、紋理特徵、彩色 圖片析示圖等。對於彩色圖片及邊緣特徵爲基礎的查詢 設計方式，最能顯示原來圖檔的資訊，以此所設計成的 CBR資料庫最爲有效。 綜合說明 經濟部智慧財產局員工消費合作社印製 本發明提出一種彩色影像資料庫建立之方法，包括 首先利用色彩空間對彩色影像作色調分析，以得到第一 索引。接著，將此彩色影像分割成數個區塊。之後，偵 測各區塊中各點之亮度値。再來，利用偵測出之各區塊 中各點之亮度値，計算出各區塊之平均數。接著，將各 區塊之平均數乘上適當比例，以分別得到對應各區塊之 平均數的門檻値。再來，依據各區塊之平均數與其各點 之亮度値，以分別得到對應各區塊之變異數。接著，比 較各區塊之門檻値與變異數，以分別得到比較結果。之 後，將這些比較結果儲存在不同之位元中，並將這些位 元作爲第二索引。最後，依據第一索引與第二索引，得 到做爲搜尋彩色影像之關係條件，以完成彩色影像資料 庫之建立。 4 本紙張尺度適用中國國家標準（CNS)A4規格（210 X 297公釐） 507452 A7 5736twf.doc/008 B7 五、發明說明（乃） 本發明提出一種彩色影像擷取之方法，包括首先輸 入色調値、亮度値及彩度値至一人機介面之點選色彩畫 面上。接著，點選此人機介面之點選物件畫面中之一區 塊。最後，依據輸入之色調値、亮度値、彩度値與點選 物件畫面中之區塊，以得到至少一彩色影像。此外，本 發明更包括設定一設定値，以判斷此彩色影像之擷取結 果，並預覽所有擷取到之彩色影像之圖形。 本發明提出之彩色影像資料庫建立的方法及彩色影 像擷取的方法，在色彩特徵是很有效率的，而且使用寬 頻的資源是非常少的，同時在速度上比一般比對 (matching)之方法可節省大量時間，對於解決第三代通 訊系統中傳輸影像最重要的兩個參數”頻寬”及”時間”有 極大的幫助。 爲讓本發明之上述目的、特徵、和優點能更明顯易 懂，下文特舉較佳實施例，並配合所附圖式，作詳細說 明如下： 圖式之簡單說明： 第1圖繪示的是資料庫建立流程圖； 經濟部智慧財產局員工消費合作社印製 (請先閱讀背面之注意事項再填寫本頁) % 第2圖是色彩空間HLS圖； 第3圖是九區分割法供邊緣特徵比對的示意圖； 第4圖是色彩空間色調分量析示圖； 第5圖繪示的是影像擷取流程圖； 第6圖是點選色彩之人機介面； 第7圖是點選物件之人機介面；以及 5 本紙張尺度適用中國國家標準（CNS)A4規格（210 X 297公釐） 經濟部智慧財產局員工消費合作社印製 507452 A7 5736twf.doc/008 β7 五、發明說明（仏） 第8圖是兩種不同圖樣供比對之圖例。 標號說明： 201 :色調 203 :亮度 205 :彩度 4 0 1 :矩形圖 601 :色調選擇棒 603 :亮度選擇棒 605 :彩度選擇棒 6 0 7 ·樣本營幕 實施例 本實施例是提供一個電子賀卡系統。彩色影像資料 庫建立在基地台，使用者透過行動電話可以在平時查閱 或每個月帳單附件中查看影像資料。當使用者想打一通” 賀卡電話”時，就可以進入查詢影像系統，將需要之圖 案以CBR的方式擷取出來。在選取需要之圖案後，可以 在空白部分輸入文字並且可以選取一段音樂（選擇性 的），之後按欲撥之電話號碼，便將此電話賀卡傳送出去。 這個電子賀卡系統是以第三代無線通訊系統傳送彩 色圖片爲對象，設計一套CBR的資料庫系統，並以視覺 化的使用者查詢介面，提供友善且快速的搜尋和擷取之 環境。 第1圖繪示的是資料庫建立流程圖。基地台開始蒐 集圖案（S101)，將蒐集到的圖案建立一個彩色影像資料 6 本紙張尺度適用中國國家標準(CNS)A4規格（210 X 297公釐） JI--1------裝·-- (請先閱讀背面之注意事項再填寫本頁) . 507452 A7 5736twf . doc/008 g-y 五、發明説明（4 ) (請先閱讀背面之注意事項再填寫本頁) 庫，這個資料庫中的圖案是經過分類，例如建築物、人、 動物及風景等（S 103)，以便日後查詢影像時可以減少很 多不必要的圖案搜尋。再將分類後的圖案做分析，判斷 這些圖案是否符合整體的效能（S105)，將一些圖案內容 .不適合或色彩不佳的予以刪除，其餘保留做爲彩色影像 資料庫中的圖案。 一般彩色影像是由RGB(Red、Green、Blue)三分量 所組成的，這三分量的關連性高而不易由其分量間求出 主色，因此本實施例是採用HLS(Hue(色調）、Lightness(亮 度）、Saturation(彩度）)色彩空間，HSI、HSV亦相同。 第2圖是色彩空間HLS圖，將HLS色彩空間量化爲色 調201的値是0°〜360°，亮度203和彩度205的値各爲 0〜100 〇 經濟部智慧財產局員工消費合作社印製 第3圖是九區分割法供邊緣特徵比對的示意圖，其 圖形爲飛機的圖形。第4圖是色彩空間色調分量析示圖。 第4圖是第3圖在執行K-means分群法而得到的矩形圖 401。在此分佈圖中，每個主色間需要一個適當的距離 (M/We = 20)，以避免使用者在選擇主色時，因色調的相因 素而導致主色選擇的錯誤以致影響後續查詢的結果（如第 1圖之步驟S107)。在第4圖中有三個主要分量，亦即三 主色，再依據三個主要分量的大小做爲主色、次主色及 次次主色，以用來當做建立彩色影像資料庫的前三個索 弓[(indices) 〇 7 本紙張尺度適用中國國家標準（CNS ) A4規格（210X297公釐） 507452 5736twf.doc/008 A7 B7 五、發明説明（L ) 在第3圖中，將圖案的影像（iV*A〇等分切成（f*f) 的區塊，再利用邊緣偵測器來做物件分析（如第1圖之步 驟S109)。先在分割的區塊中求得一個平均數，用區塊 (f*f)各點的亮度値相加得其總和，再用其總和除以區 塊（f*f)之總點數而得到這個平均數。其次，在分割的 區塊中求得一個變異數，用區塊各點的亮度値減 經濟部智慧財產局員工消費合作社印製 掉平均數，其各點的差額平方後再相加而得到總和，這 個總和即是變異數。設定一個門檻値（Threshold)來判斷 該偵測的區塊是否有物件存在，一般門檻値爲該區塊之 平均數的75%。 用9個位元（bits)記錄9個區塊中是否有物件存在， 區塊中有物件存在則相對應之位元記錄”1”，反之則記 錄”〇”。當每一個區塊的變異數大於門檻値時，表示區塊 中有物件並記錄”1”於位元之中，反之則記錄”0”。因此， 依序完成上述之工作並記錄這9個位元，將這9個位元 作爲建立彩色影像資料庫的第四個索引。 將蒐集的圖案經分類整理後（如第1圖之步驟 S101)，由色彩空間之色調分析找出三主色（如第1圖之 步驟S107)，當做建立彩色影像資料庫的前三個索引， 再用邊緣偵測器偵測區塊中的物件（如第1圖之步驟 S109)，所記錄的9個位元做爲第四個索引。這四個索引 爲搜尋彩色影像的條件，將此關係條件編碼而成爲該彩 色影像之檔名，即完成彩色影像資料庫的建立（如第1圖 8 (請先閱讀背面之注意事項再填寫本頁) •裝. ,ιτ 本紙張尺度適用中國國家標準（CNS ) A4規格（210X297公釐） 經濟部智慧財產局員工消費合作社印製 507452 5736twf.doc/008 〜 B7 五、發明説明（7) 之步驟S 111)。 第5圖繪示的是影像擷取流程圖。在本實施例中， 使用者利用行動電話打一通”賀卡電話”，就可以進入查 詢影像系統（S501)。 第6圖是點選色彩之人機介面。此時，行動電話螢 幕會出現請輸入第一主色的畫面（如第5圖之步驟 S5〇3) ’畫面中會出現Hue選擇棒601、Saturation選擇 棒603及Lightness選擇棒605之三個選擇棒。在Hue 選擇棒601選出所要的顏色，而Saturation選擇棒603 及Lightness選擇棒6〇5可以選擇性（option)使用。這時， Sample螢幕607會顯現由Hue選擇棒601、Saturation 選擇棒603及Lightness選擇棒605三個選擇棒所選擇 的結果。依照上述的步驟，同樣完成第二主色和第三主 色的輸入。 第7圖是點選物件之人機介面。當完成上述三個步 驟時，行動電話螢幕會出現請輸入物件所在方塊的畫面 (如第5圖之步驟S505)，畫面中有9個方塊，所點選的 方塊即是物件所在之處。 當三主色及物件所在方塊輸入完畢之後，首先，以 輸入的三主色的資料做爲搜尋的資料，比對彩色影像資 料庫的前三個索引，若其搜尋結果相同，則將搜尋到的 彩色影像擷取出來（如第5圖之步驟S507)。^主色的資 料與彩色影像資料庫內前三個索引的資料逐〜比對。當 搜尋的結果低於設定値（例如十張彩色影像）時，將捜尋 9 本紙張尺度適用中( CNS ) Λ4規格（210><297公釐1 " '^ - (請先閲讀背面之注意事項再填寫本頁) 裝. 訂 507452 A7 B7 5736twf#cj〇c/QQg 五、發明説明（2 ) 到所有的彩色影像，每一張都在螢幕上預覽，供使用者 選定其所要的圖形（如第5圖之步驟S509)。 (請先閱讀背面之注意事項再填寫本頁) 若搜尋的結果高於設定値時，再將輸入物件所在方 塊的資料做爲搜尋的資料，並以輸入三主色做爲搜尋的 .資料已經擷取出來的彩色影像再做一次的搜尋。點選物 件所在方塊的位置與彩色影像資料庫內第四個索引之9 個位元的位置逐一比對。若其搜尋結果相同，將搜尋到 的彩色影像擷取出來（如第5圖之步驟S507)。將搜尋到 所有的彩色影像，每一張都在螢幕上預覽，供使用者選 定其所要的圖形（如第5圖之步驟S509)。 第8圖是兩種不同圖樣供比對之圖例。這兩種不同 的圖樣經色調分量析示後，發現其三主色皆相同。在經 過邊緣偵測器的辨識後，知道物件在區塊中的分布是不 相同的。如此，使用者可以減少預覽圖形的數量。 經濟部智慧財產局員工消費合作社印製 本實施例的彩色影像資料檔爲100張的Corel Draw 7.0 Photo彩色影像，將影像大小調整爲255*255*3Bit (24 Bit True color)。利用彩色影像資料庫中各色彩影像 經分析分類、色彩空間轉換與邊緣偵測器（edge detector) 後，依其關係特性編碼建立檔案。 在彩色影像查詢的部分，使用者先選取分類影像’ 其次再選擇三主色，最後再點選物件所在的方塊，螢幕 上預覽所有搜尋到的彩色影像。結果在第一次選取的主 色其篩選率爲26%(相對於本資料庫之影像）。.在第二次 選取的次主色其篩選率爲8%(相對於本資料庫之影像）。 10 本紙張尺度適用中國國家榡準（CNS〉A4规格（210X297公釐） 507452 A7 5736twf . doc/008 ... 五、發明説明（q ) 在第三次選取的次次主色其篩選率爲7%(相對於本資料 庫之影像）。若再加上邊緣偵測器的篩選其篩選率降至2% (相對於本資料庫之影像）以下。 綜上所述，雖然本發明已以較佳實施例揭露如上， .然其並非用以限定本發明，任何熟習此技藝者，在不脫 離本發明之精神和範圍內，當可作各種之更動與潤飾， 因此本發明之保護範圍當視後附之申請專利範圍所界定 者爲準。 (請先閲讀背面之注意事項再填寫本頁) 經濟部智慧財產局員工消費合作社印製 本紙張尺度適用中國國家標準（CNS ) A4規格（210X 297公釐）507452 A7 5736twf.doc / 008 R7 Description of the invention (() Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, printed the field of invention (please read the precautions on the back before filling this page) This invention is about a color image database Method and method for color image capture, and in particular, it relates to a method for establishing a color image database and a method for color image capture which are of great help in transmitting images in a communication system. BACKGROUND OF THE INVENTION The digital world is full of various formats In the storage and processing of data and media, there must be many mechanisms such as code retrieval and query to meet general or real-time retrieval. Among them, Content-Based Retrival (CBR) is the most important solution. The connotation of various data or media includes digital text, sound effects, music, pictures, animation movies, etc., to build a database, which can improve the performance of these data signal storage management systems, because the traditional database system is usually based on the data itself Simple attributes to create or search, create, update and check in management In addition, CBR can also provide a similar mechanism to browse the media in the widely used World Wide Web today, provide search engine search index, and some people in the World Wide Web search engine to provide keywords for web document queries Although the design is simple and fast, the query of keywords has the following disadvantages: (1). Can not handle some query languages that are not easy to describe, for example, "find photos of landscapes with the sun." (2). Key The wrong words are used to seriously affect the results of ingestion. (3). The description method is a permutation and combination of some single keywords, which results in inaccuracy and incompleteness of the query. (4). Keyword 3 This paper scale applies to China National Standard (CNS) A4 Specification (210 X 297 mm) 507452 A7 5736twf.doc / 008 B7 V. Invention Description (1) The database needs to be classified by people in advance, and the database cannot be created automatically. (Please read the back first Please note this page before filling in this page) If all kinds of data or media use CBR to establish a storage management system and complete the query, the data itself can be enhanced. Like pictures, it is more necessary to use CBR to achieve effective storage and ingestion functions. In CBR's research on image pictures, there are many ways to complete the query. As shown above, keywords, image edges, texture features, and color picture analysis For example, for the query design method based on color pictures and edge features, the information of the original image file can be displayed most effectively, and the CBR database designed by this is the most effective. The invention proposes a method for establishing a color image database, which includes firstly performing color tone analysis on a color image using a color space to obtain a first index. Then, the color image is divided into several blocks. After that, the brightness of each point in each block is detected. Then, the average number of each block is calculated using the detected brightness 値 of each point in each block. Next, multiply the average number of each block by an appropriate ratio to obtain the thresholds 値 corresponding to the average number of each block. Then, according to the average number of each block and the brightness of each point, to obtain the variation number corresponding to each block. Then, the threshold 値 and variation number of each block are compared to obtain the comparison results. Thereafter, these comparison results are stored in different bits, and these bits are used as a second index. Finally, according to the first index and the second index, the relational conditions for searching color images are obtained to complete the establishment of the color image database. 4 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 507452 A7 5736twf.doc / 008 B7 V. Description of the invention (yet) The present invention proposes a method for color image capture, including first inputting the hue値, Brightness 彩 and Saturation 値 to the click color screen of a terminal. Then, click a block in the Click Object screen of this terminal. Finally, at least one color image is obtained according to the input hue 値, brightness 値, chroma 値 and clicking on a block in the object screen. In addition, the present invention further includes setting a setting frame to judge the capture result of the color image and preview all the captured color image patterns. The method for establishing a color image database and the method for capturing color images proposed by the present invention are very efficient in color characteristics, and use very few broadband resources. At the same time, they are faster than ordinary matching The method can save a lot of time, and it is of great help to solve the two most important parameters of "transmission bandwidth" and "time" in the third-generation communication system. In order to make the above-mentioned objects, features, and advantages of the present invention more comprehensible, the preferred embodiments are described below in detail with the accompanying drawings as follows: A brief description of the drawings: FIG. 1 illustrates It is the flow chart for the establishment of the database; it is printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs (please read the notes on the back before filling this page)% Figure 2 is the color space HLS map; Figure 3 is the nine-zone division method for the edge Schematic diagram of feature comparison; Figure 4 is a color space tonal component analysis chart; Figure 5 is a flowchart of image capture; Figure 6 is a human-machine interface for selecting colors; Figure 7 is a selected object Human-machine interface; and 5 paper sizes applicable to China National Standard (CNS) A4 (210 X 297 mm) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 507452 A7 5736twf.doc / 008 β7 5. Description of the invention (仏) Figure 8 is an illustration of two different patterns for comparison. Explanation of symbols: 201: hue 203: brightness 205: chroma 4 0 1: rectangular chart 601: hue selection bar 603: brightness selection bar 605: chroma selection bar 6 0 7 Electronic greeting card system. The color image database is built on the base station, and users can view the image data in their mobile phones or in the monthly bill attachment. When the user wants to make a "greeting card call", he can enter the query image system and retrieve the required images in CBR mode. After selecting the desired pattern, you can enter text in the blank part and select a piece of music (optional), and then press the phone number you want to dial to send this phone greeting card. This electronic greeting card system is designed to transmit color pictures in the third generation wireless communication system. A CBR database system is designed, and a visual user query interface is provided to provide a friendly and fast search and retrieval environment. Figure 1 shows a flowchart of database creation. The base station starts to collect patterns (S101), and creates a color image data from the collected patterns. 6 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm). JI--1 ------ installation ·-(Please read the notes on the back before filling this page). 507452 A7 5736twf. Doc / 008 gy V. Description of the invention (4) (Please read the notes on the back before filling this page) Library, this database The patterns in it are classified, such as buildings, people, animals, and landscapes (S 103), so that many unnecessary pattern searches can be reduced when querying images in the future. Then analyze the classified patterns to determine whether these patterns meet the overall performance (S105), delete some pattern contents that are not suitable or have poor colors, and leave the rest as patterns in the color image database. Generally, a color image is composed of three components of RGB (Red, Green, and Blue). These three components are highly related and it is not easy to find the main color between the components. Therefore, in this embodiment, HLS (Hue (hue), Lightness (Saturation) color space, HSI, HSV are the same. Figure 2 is a color space HLS diagram. The HLS color space is quantized into hue 201, which is 0 ° ~ 360 °, and brightness 203 and chroma 205, which are each 0 ~ 100. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economy Figure 3 is a schematic diagram of edge feature comparison for the nine-region segmentation method, and its graph is the graph of an airplane. Fig. 4 is a color space hue component analysis diagram. Fig. 4 is a histogram 401 obtained by executing the K-means clustering method in Fig. 3. In this distribution map, an appropriate distance is required between each main color (M / We = 20) to avoid the user's selection of the main color, which will lead to the wrong selection of the main color due to the phase of the hue, which will affect subsequent queries. (As in step S107 in FIG. 1). In Figure 4, there are three main components, that is, the three main colors, and the main color, sub-primary color, and sub-primary color are based on the size of the three main components, which are used as the first three to build the color image database. Individual bows [(indices) 〇7 This paper size applies to the Chinese National Standard (CNS) A4 specification (210X297 mm) 507452 5736twf.doc / 008 A7 B7 V. Description of the invention (L) In Figure 3, The image (iV * A〇 is equally divided into (f * f) blocks, and then an edge detector is used for object analysis (such as step S109 in FIG. 1). Firstly, an average is obtained in the divided blocks. The average value is obtained by adding the brightness 区块 of each point of the block (f * f), and then dividing the sum by the total number of points of the block (f * f) to obtain this average. Second, in the divided area Obtain a variation number in the block. Use the brightness of each point in the block to reduce the average number printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. Add the squared differences and add up the sum. This sum is the variation Set a threshold to determine whether there are objects in the detected block. Generally The threshold is 75% of the average number of the block. Use 9 bits to record whether there are objects in 9 blocks. If there are objects in the block, the corresponding bit record is "1", and vice versa. Then record "0". When the variation number of each block is greater than the threshold 値, it means that there are objects in the block and record "1" in the bit, otherwise record "0". Therefore, complete the above in order Work and record these 9 bits, and use these 9 bits as the fourth index to build the color image database. After sorting the collected patterns (such as step S101 in Figure 1), the hue of the color space is determined. Analyze and find the three main colors (as in step S107 in Fig. 1), and use them as the first three indexes to build the color image database, and then use the edge detector to detect objects in the block (as in step S109 in Fig. 1) The recorded 9 bits are used as the fourth index. These four indexes are the conditions for searching for color images, and this relationship condition is encoded to become the file name of the color image. This completes the establishment of the color image database (such as Figure 1 Figure 8 (Please read the notes on the back before filling out this page) • Package., Ιτ This paper size is applicable to Chinese National Standard (CNS) A4 (210X297 mm) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 507452 5736twf.doc / 008 ~ B7 V. Step S of Invention Description 111). Figure 5 shows a flowchart of image capture. In this embodiment, the user can enter the query image system (S501) by making a "greeting card call" using a mobile phone. Figure 6 is a click Color human-machine interface. At this time, the screen of the mobile phone will appear. Please enter the first main color (as shown in step S503 of Figure 5). 'The Hue selection bar 601, Saturation selection bar 603, and Lightness selection will appear on the screen. Three options for Stick 605. The desired color is selected in Hue selection bar 601, and Saturation selection bar 603 and Lightness selection bar 605 can be used selectively. At this time, the Sample screen 607 will show the result selected by the three selection bars Hue selection bar 601, Saturation selection bar 603 and Lightness selection bar 605. Follow the above steps to complete the input of the second and third main colors. Figure 7 is the man-machine interface for clicking objects. When the above three steps are completed, the screen of the mobile phone will appear, please enter the box where the object is located (such as step S505 in Figure 5). There are 9 boxes in the screen, and the selected box is where the object is located. After the input of the three main colors and the box where the object is located, first, use the input data of the three main colors as the search data, and compare the first three indexes of the color image database. If the search results are the same, you will find The color image is captured (as shown in step S507 in FIG. 5). ^ The main color data is compared with the data of the first three indexes in the color image database. When the search result is lower than the setting (for example, ten color images), 9 paper sizes are applicable (CNS) Λ4 specification (210 > < 297 mm1 " '^-(Please read the back first Please note this page before filling in this page) Binding. Order 507452 A7 B7 5736twf # cj〇c / QQg V. Description of the invention (2) To all the color images, each one is previewed on the screen for users to choose what they want Graphic (such as step S509 in Figure 5). (Please read the precautions on the back before filling in this page) If the search result is higher than the setting, use the data in the box where the input object is located as the search data, and use the Enter the three main colors as the search. Search the color image that the data has been extracted for. Search the position of the box where the object is located and compare it with the 9-bit position of the fourth index in the color image database. If the search results are the same, extract the color images that have been searched (such as step S507 in Figure 5). All the color images will be searched, and each one will be previewed on the screen for the user to select the desired one. Graphic (as in Figure 5) S509). Figure 8 is an example of two different patterns for comparison. After analyzing the hue components of the two different patterns, it is found that the three main colors are the same. After recognition by the edge detector, the object is known The distribution in the blocks is different. In this way, users can reduce the number of preview graphics. The Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs printed 100 color image data files of this embodiment with Corel Draw 7.0 Photo color images , Adjust the image size to 255 * 255 * 3Bit (24 Bit True color). Use each color image in the color image database to analyze and classify, color space conversion and edge detector, and then encode according to its relationship characteristics Create a file. In the part of the color image query, the user first selects the classified image ', then selects the three main colors, and finally clicks the box where the object is located to preview all the searched color images on the screen. The primary color has a screening rate of 26% (relative to the image in this database). The secondary primary color selected in the second selection has a screening rate of 8% (relative to the image in this database) 10 This paper size applies to China National Standard (CNS> A4 size (210X297 mm) 507452 A7 5736twf .doc / 008 ... 5. Description of the invention (q) The third primary color selected in the third time has a screening rate of 7% (relative to the image of this database). If the edge detector is added, the screening rate will be reduced to less than 2% (relative to the image of this database). In summary, although the present invention has been The preferred embodiment is disclosed as above. However, it is not intended to limit the present invention. Any person skilled in the art can make various changes and decorations without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be Subject to the scope of the attached patent application. (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs This paper is sized for the Chinese National Standard (CNS) A4 (210X 297 mm)