1294100 (1) 九、發明說明 【發明所屬之技術領域】 本發明係關於攜帶型資訊終端之光學式文字辨識時之 處理對象選擇方法。 【先前技術】 如攝影機等配載著影像輸入機能之攜帶型資訊終端， 正積極開發針對攝取之影像實施光學式文字辨識之技術。 然而，攜帶型資訊終端之攝影機所攝取之影像因爲手晃動 等原因，攝取之文字列往往相對於畫面會呈傾斜。因此， 文字辨識結果錯誤（誤讀）時，使用者必須調整攝取時之角 度再度攝取、或利用鍵盤等輸入裝置修正辨識結果。 實施辨識處理之前用以補正辨識對象文字列之傾斜之 技術如專利文獻1所示，係檢測預先記載於記載著辨識對 象文字列之用紙之2個以上之標記，並對應以此方式得到 之傾斜自動實施旋轉後再進行辨識。 又，非專利文獻1之攜帶型資訊終端之英單字辨識係 結合查閱英日辭典之文字辨識、及辭典檢索之方法。 [專利文獻1]日本特開平1卜25 0 1 79號公報（4〜7項、第 3圖） [非專利文獻 l]H.Fujisawa，H.Sako，Y.Okada，and S-W.Lee,u Information Capturing Camera and Developmental Issues，’’In Proc.Int.Conf. Document Analysis and Recognition,ICDAR799,Bangalore, India,Sep.20-22,1999, 1294100 (2) pp.205-208 0 · 實施良好精度之文字辨識處理上，影像内之辨識對象 · 之文字列沒有傾斜是極爲重要的。然而，利用數位攝影機 及附有攝影機之攜帶型終端等資訊終端裝置攝取數位影像 時，通常是手持該資訊終端裝置，要使辨識對象之文字列 在影像内不會呈傾斜，在攝取時必要特別注意（第一課題） 〇 針對以附攝影機能之攜帶型終端攝取之影像之光學式 · 文字辨識機能，在將文字列拍攝成傾斜時，傳統技術有無 法執行文字辨識、或無法具有充分辨識精度之問題。因此 ，有必須重新攝取辨識對象之文字才能得到正確之文字辨 識結果之課題。 專利文獻1記載著利用記載於用紙上之2個以上之標記 來檢測傾斜之內容，然而，該技術必須預先在辨識對象之 用紙上印製或記入標記，在辨識名片等時並不實用。又， 針對過去所攝取之影像實施文字辨識處理時，若未能如上 鲁 述所述在特別注意下進行拍攝，則辨識對象文字列即可能 在影像内呈現傾斜。此時，在實施文字辨識處理前，必須 利用影像處理軟體等實施影像角度之補正（第二課題）。 此外，文字之記述方向有直書及橫書之2種時，即使 調整影像角度來抽出辨識對象文字列，在實施文字辨識時 ，必須設定辨識對象文字列爲直書或橫書。取得名片等之 影像並實施文字辨識處理時，因爲已存在直書或橫書之格 式，故有必須每次設定辨識模式之問題。此外，報紙、雜 -6- 1294100 (3) 誌等混合存在直書文字列及橫書文字列之文書時，切換文 > 字之記述方向係很大的負擔（第三課題）。 · 利用附有攝影機之攜帶型資訊終端等之裝置時，經常 會在運輸工具移動中等無法在安定位置攝取影像之環境下 使用。然而，傳統技術時，辨識對象文字列相對於影像之 傾斜若未能選擇抑制於較小之狀態，則無法獲得精度良好 之文字辨識。因此，使用者在攝取辨識對象影像時，必須 特別注意角度調整（第四課題）。 修 此外，利用附有攝影機之攜帶型資訊終端之文字辨識 結果實施辭典檢索及網際網路之檢索時，英單字係以空白 來區分單字單位，故檢索對象之單字選擇較爲容易。然而 ，對日文實施同樣處理時，則和英單字時不同，因爲並不 存在利用單字間之空白之布置資訊來進行區隔，使用者必 須針對每1文字選擇檢索對象之文字列，故指定上會產生 相當大的負擔（第五課題）。 【發明內容】 有鑑於上述課題，本發明之目的即在提供一種系統及 方法，可減輕利用附攝影機能之攜帶型資訊終端攝取文字 辨識對象物時之攝取者負擔、及可減輕將攝取後之影像補 正成適合文字辨識之角度時之負擔。 此外，本發明之目的在即供一種系統及方法，在實施 日文之文字辨識並實施辭典及網際網路之檢索時，可減輕 使用者指定檢索對象之文字列之負擔。 1294100 ' (4) 爲了解決上述第一課題’在資訊終端裝置之畫面上即 時顯示可將辨識對象文字列之影像内之傾斜程度呈現給攝 β 取者之角度指示器。使用者觀察角度指示器顯示之資訊’ 移動資訊終端裝置或辨識對象物’在適當之位置進行拍攝 ，可攝取適合文字辨識處理之影像。 爲了解決上述第二課題，提供一種機能，旋轉已攝取 之影像之同時，即時實施辨識對象文字列之行抽出。使用 者利用簡單之按鍵操作等，在資訊終端裝置之畫面上旋轉 · 已攝取影像，並確認即時執行行抽出之辨識對象文字列候 補。在抽出期望之辨識對象文字列之時點，使用者停止旋 轉處理，抽出複數行時，選擇辨識文字列並執行辨識處理 ，可簡化針對過去所攝取之影像之文字辨識處理。 爲了解決上述第三課題，提供自動判定該辨識對象文 字列係直書或橫書之機能。該辨識對象文字列方向之自動 判定上，係利用抽出之辨識對象文字列之外接矩形之長寬 比。具體而言，將辨識對象文字列之外接矩形之高及寬之 · 比率與規定値進行比較，判斷係直書文字列或橫書文字列 ，並實施辨識處理。使用之攜帶型終端裝置之畫面方向爲 可變時，亦可自動實施對應畫面方向之直書及橫書之切換 〇 爲了解決上述第四課題，提供用以產生依據辨識對象 文字列之傾斜角度自動選擇辨識對象文字列之外接矩形之 手段。具體而言，利用例如日本特開平7- 1 4 1 465「文書影 像之傾斜檢測方法」之方法檢測辨識對象文字列之傾斜角 -8- 1294100 (5) 度，辨識對象文字列爲橫書時建立將影像朝水平位置旋轉 ' 之影像、直書時建立將影像朝垂直位置旋轉之影像。其後 < ，產生辨識對象文字列之外接矩形，建立追加至該旋轉後 影像之旋轉外接矩形影像。接著，旋轉該旋轉外接矩形影 像使其回到基準之文字列之傾斜角度，並顯示於資訊終端 裝置之顯示裝置。 爲了解決上述第五課題，提供一種手段，針對日文之 文字辨識結果執行構詞分析，自動產生檢索對象之文字列 · 候補，由使用者選擇各候補、或擇各文字列候補之組合。 以視覺容易了解之方式顯示辨識對象之影像之傾斜情 形，可使使用者得知影像係傾斜狀態，而容易以較適合文 字辨識處理之角度攝取影像。 又，過去已攝取之影像方面，係直接編輯傾斜之影像 而可實施文字辨識處理，故無需再度拍攝。 又，因爲角度補正困難時，亦可以傾斜狀態來選擇文 字列，故對具某種程度傾斜之影像亦可執行文字辨識處理 鲁 【實施方式】 參照第1〜20圖，針對本發明良好實施形態進行說明 。本發明之文字辨識方式可應用於讀取並辨識名片等之移 動資訊系統等，例如，亦可應用於行動電話之文字辨識機 第1圖係應用本發明之攜帶型資訊終端之方塊圖之實 -9 - 1294100 (6) 例。本實例時，具有：攜帶型資訊終端本體1 00 ;用以光 學輸入辨識對象之名片等之攝影機及掃描器等之影像輸入 裝置1 1 0 ;用以顯示辨識對象之影像、文字辨識結果、以 及游標121等之CRT及液晶等之顯示裝置120 ;配置著使 用者操作按鈕1 3 1等之輸入裝置1 3 0 ;以執行攜帶型資訊終 端之整體控制爲目的而配載於終端機本體1 〇〇内之控制部 140及文字行抽出部151 ;用以執行文字辨識處理152等之 文字辨識部1 5 0 ;以及具有文字行傾斜之量化機能1 6 1及影 像旋轉處理機能162之影像處理部160。 文字辨識部150及影像處理部160亦可以爲軟體之機能 ，只要在和控制部1 40相同之電路上動作即可。輸入裝置 爲按鈕等一般裝置即可，然而，爲了提高操作性，亦可以 觸摸面板等之顯示裝置兼輸入裝置來實現顯示裝置120及 輸入裝置1 3 0。 第2圖係實施應用以解決第一課題爲目的之手段之文 字辨識處理時之流程圖之實例（第一實施形態）。使用者開 始執行OCR機能時，從影像輸入裝置1 1〇輸入之移動影像 會顯示於顯示裝置1 2 0 ( S 2 0 1)。文字辨識部1 5 0立即針對顯 示於影像顯示裝置120之十字型等之游標121所在之區域執 行文字行抽出處理（S2 02)，並顯示用以框取抽出文字行之 外接矩形（S203 )。 同時，利用影像處理部1 60執行文字列之傾斜之量化 ’將經過量化之値以條形圖等之形式顯示於角度指示器 1 23 (S204)。經過量化之値只要可以反映文字列及影像之 1294100 (7) 傾斜程度即可，例如，採用與文字行及影像邊緣所形成之 角度Θ之倒數成正比之値，傾斜較小時，角度指示器1 2 3會 顯示較大之値，使用者可以利用直覺執行操作。 至使用者按下攝取按鈕（S20 5 )爲止，會回到步驟 (S201)之處理，並重複以後（S201〜S204)之處理，持續更 新畫面顯示。 使用者參考角度指示器1 23並移動終端裝置或辨識對 象物，在適當攝取位置按下攝取按鈕（S20 5)即可執行影像 攝取（S206)。其次，按下辨識按鈕（S207)時，會對步驟 (S203 )顯示之外接矩形内之文字列執行文字辨識處理210 並顯示辨識結果（S211)。 按下攝取按鈕（S20 5 )後，若按下取消按鈕（S208)，則 會捨棄攝取之影像並回到步驟（S201)。按下其他按鈕時， 執行相關必要動作（S209)。 第3圖係實施應用以解決第二課題爲目的之手段之文 字辨識處理時之流程圖之實例（第二實施形態）。攜帶型資 訊終端之使用者執行以前攝取之靜止影像之載入（S301)， 該靜止影像會被顯示於顯示裝置120(S 3 02)。文字辨識部 150立即執行文字行抽出（S 3 03 )、外接矩形顯示（S 3 04)、以 及角度指示器顯示（S 3 0 5 )。在此時點，處理會進入等候使 用者之按鍵輸入（S 3 06)之階段。 使用者按下辨識執行按鈕（S 3 07)時，立即針對在步驟 (S 3 04)顯示之外接矩形之内部之文字列執行文字辨識處理 (S3 10)，並顯示文字辨識結果（S3 11)。使用者按下箭頭按 1294100 (8) 鈕（s 3 0 8 )時，利用按下之按鈕使影像朝右或左之方向旋轉 一定角度（S3 09)。此時，旋轉中心係文字行矩形之中心， 然而’亦可以使用者指定之影像上之一點爲中心來執行旋 轉處理。 持續按住箭頭按鈕時，影像會連續旋轉，且外接矩形 顯示亦會即時隨之更新。若考慮使用者之便利性，亦可以 停止箭頭按鈕按下（手指離開按鈕）之動作來取代辨識執行 按鈕按下之操作。 第4圖係第一實施形態及第二實施形態時，顯示於顯 示部120之影像實例。 第一實施形態時，顯示部120會顯示影像輸入裝置110 輸入之動畫，第二實施形態時，則顯示部1 2 0會顯示以前 攝取之靜畫400。 此時點，影像4 0 0内之辨識對象文字列4 0 1係呈現傾斜 狀態。畫面中央顯示著旋轉中心之十字游標4 0 2。文字辨 識部150產生用以框取辨識對象文字列401之外接矩形403 ，並顯示於畫面上。角度指示器顯示用以表示傾斜之辨識 對象文字列401之傾斜程度之條形圖（404)。 第一實施形態時，係以旋轉手持之攜帶型資訊終端（ 攝影機）（4〇6)來旋轉顯示於畫面上之影像，又，第二實施 形態時，則係操作箭頭鍵（405)來旋轉顯示於畫面上之影 像。辨識對象文字列40 1之傾斜會隨著影像之旋轉而變小 ，而外接矩形亦會產生同步之形狀變化（407)。 此外，角度指示器利用條形圖顯示代表辨識對象文字 -12- 1294100 Ο) 列401之傾斜變小之較大的値（408)。利用者重複執行影像 之旋轉操作，在角度指示器顯示之値爲較大之位置執行文 字辨識處理，可得到高精度之文字辨識結果。 第5圖係實施應用以解決第三課題爲目的之手段之文 字辨識處理時之流程圖之實例（第三實施形態）。步驟 (S501〜S 5 06)與步驟（S201〜S209)相同，故省略其說明。 辨識按鈕按下（S 5 04)時，計算辨識對象文字列之外接 矩形之長寬比（高及寬之比率），將其與預先規定之値α進 行比較（S5 07)。長寬比大於規定値α時，判斷成直書文字 列’實施直書文字列辨識用之參數設定等（S 510)，執行文 字辨識處理（S51 1)並顯示結果（S512)。 同樣的，若長寬比不大於規定値α，接著，將長寬比 與規定値/3進行比較（S 5 08)。長寬比小於規定値Θ時，判 斷成橫書文字列，實施橫書文字列辨識用之參數設定等 (S5 09)，執行文字辨識處理（S51 1)並顯示結果（S512)。長 寬比爲α以下、Θ以上之範圍時，則被視爲不對文字列實 施充分角度補正而不執行辨識處理。 第6圖係第三實施形態之顯示於顯示部1 20之影像實例 。橫書之文字列601時，外接矩形之高602會小於寬603。 若以外接矩形之高/外接矩形之寬來定義長寬比，長寬比 小於1時，外接矩形係橫長。 例如，將用以和長寬比進行比較之規定値Θ設定成 〇 . 5時，在長寬比低於〇 . 5之狀態啓動文字辨識，會自動實 施橫書文字列之必要設定。同樣的，直書文字列604時， 1294100 (10) 外接矩形之高605會大於寬606。 若爲和上述相同之長寬比之定義，則長寬比大於1時 ，外接矩形爲縱長。例如，將用以和長寬比進行比較之規 定値α設定成1 · 5時，在長寬比高於丨.5之狀態啓動文字辨 識，會自動實施直書文字列之必要設定。 第7圖係實施應用以解決第四課題爲目的之手段之文 字辨識處理時之流程圖之實例（第四實施例）。 從影像輸入裝置11 0輸入文字辨識對象影像（S 7 0 1)時 ，影像處理部1 60會立即計算辨識對象文字列相對於影像 之角度（S 702)，對輸入之文字辨識對象影像實施該角度份 旋轉補正。補正方向方面，橫書文字列時爲朝畫面之水平 方向旋轉，直書文字列時爲朝畫面之垂直方向旋轉。 其次，針對該旋轉後影像實施文字列抽出，對該旋轉 後影像附加外接矩形（S 704)，將該影像儲存於影像緩衝器 。儲存於影像緩衝器之影像可以爲影像整體，亦可以只爲 附加之外接矩形之内部。 接著’以步驟（S702)檢測到之角度份實施該旋轉後影 像之逆旋轉，建立和原來輸入影像同樣傾斜之影像，並顯 示於終端裝置之顯示部1 2 0。使用者若執行任何操作時， 回到步驟（S 701)，針對新的輸入影像重複執行步驟（S 701 〜S707)之處理。 按下辨識按鈕時，讀取在步驟（S 7 0 5 )儲存至影像緩衝 器之影像（S 710)，針對該影像執行文字辨識處理（S71 1)並 顯示文字辨識結果（S 7 1 2 )。 1294100 (11) 第8圖係第四實施形態處理之影像狀態之實例。從影 像輸入裝置1 10輸入之辨識對象影像801上之辨識對象文字 列8 02對畫面呈傾斜狀態。影像處理部丨6〇檢測到該辨識對 象文字列802和畫面邊緣之角度8〇3，將影像旋轉角度803 ’將辨識對象文字列補正成對畫面成水平，建立旋轉補正 影像804。 文字辨識部150針對旋轉補正影像8 04上之辨識對象文 字列806實施文字行抽出，並附加外接矩形806。影像處理 部160將該旋轉補正影像8〇4實施角度803之逆旋轉，產生 回到與原來之辨識對象影像8 0 1相同之角度之影像8 0 7，並 顯示於顯示裝置120。 以上之第一至第四實施形態可以各別實現，亦可以選 擇全部或其中一部份來實現。 其次，參照第9圖至第1 2圖，針對用以解決第5課題之 第5實施形態進行說明。第9圖係應用第5實施形態之攜帶 型資訊終端之方塊圖之實例。和第1圖之差異，係追加了 檢索語抽出部170及電子辭典171。 第1 〇圖係在第9圖之裝置實施利用解決第5課題之手段 之文字辨識及辭典檢索處理時之流程圖之實例。從影像輸 入裝置1 10輸入文字辨識對象影像（S1001)，影像處理部 160對來源影像實施適當影像處理後，文字辨識部150之文 字行抽出部1 5 1會抽出文字行（S 1 0 0 2 )。其後，文字行辨識 部1 52針對抽出之文字行進行文字切割並輸出每丨文字之辨 識結果（S 1 003 )。辨識結果包括每1文字之文字碼及對應之 1294100 (12) 矩形座標。 將文字辨識之結果輸入檢索語候補抽出部1 7 0 ’並利 用構詞分析將連續之文字列分解成單字等之構詞（S 1 〇 0 4) 。例如，顯示於第9圖之攜帶型資訊終端之顯示器之「文 法Φ知識$•使？ &」之文字列時，會產生如第1 1圖所示之 表1 1 00。儲存著利用構詞分析實施分解之文字列、及對應 之矩形座標。 利用表1 1 〇 〇之資料在攜帶型資訊終端之顯示器顯示檢 索對象之候補語（S 1 005)。例如，將游標之中心座標及表 1 1 〇〇之候補語矩形座標進行比較，將含有游標之中心座標 之候補語之矩形座標以第12圖之1 200所示之形式顯示於顯 示器。 其次，按下游標移動按鍵174或176，可以如1201之方 式移動檢索對象語之矩形顯示，顯示想要檢索之矩形表時 ’按下選擇按鈕175來決定檢索對象語（S1006)。將決定之 檢索對象語輸出至電子辭典部1 7 1。電子辭典部1 7 1則將輸 入之檢索對象語視爲關鍵字實施電子辭典之檢索（S1 〇〇7) ’並將檢索結果顯示於顯示器（S 1 00 8)。 本實施例係利用構詞分析來產生檢索對象之候補語， 然而’亦可以採用將辨識結果之文字列中之「漢字」變換 成「平假名」等變更文字之種類時進行區隔並產生候補語 之方法。或者，亦可在文字種類、空白、以及文字尺寸改 變之位置等配合幾何學資訊來進行判斷。 其次’參照第9圖、第10圖、以及第13圖至第15圖， -16- 1294100 (13) 針對以解決第5課題爲目的之第6實施形態進行說明。本實 < 施例時，係如「…Φ文法知識全使）」…」所示，以含有 * 由複數構詞所構成之漢字列時爲例，且假設游標位於「文 法知識」之文字列之其中之一。 使用者可能只想以「文法」或「知識」之構詞做爲檢 索對象、或以「文法知識」之複合語整體做爲檢索對象’ 因此，以下針對減輕此選擇之負擔之處理進行說明。因爲 第6實施形態與第1 0圖之處理流程不同之部位’係候補語 鲁 抽出（S 1 004)、候補語顯示（S 1 005 )、以及候補語選擇 (S 1 006)，故只針對前述處理進行說明。 候補語抽出（S 1 00 4)時，係以和第5實施例所示相同之 處理利用構詞分析來產生候補語，並產生對應其之第1 3圖 所示之矩形座標表1 3 00。其次，產生以從表1 3 00分別選擇 「文法」、「文法知識」、以及「知識」爲目的之區域座 標表1 400。該表係游標中心座標進入該區域時以顯示對應 其之候補語之矩形爲目的之表。 ® 第1 5圖係該表之矩形座標之X座標之模式圖。分別 對應於「文法」、「文法知識」、以及「知識」之選擇區 域係1 5 0 0、1 5 0 1、以及1 5 0 2，依據游標之中心座標位於那 —區域而從表1 400選擇用以顯示檢索候補語之矩形座標。 選取之矩形座標會顯示於顯示器（S 1〇〇5)。 第1 6圖係顯示之矩形實例。（a)係游標位於「知識」 之選擇區域時之顯示例，（b)及（c)則係游標分別位於「文 法知識」及「文法」之選擇區域時之顯示例。使用者在顯 -17- 1294100 (14) 示著欲檢索之候補語之矩形時，按下選擇按鍵1 7 5即可選 取檢索對象語（S 1 006)。本實施例係針對1文字行内之文字 _ 列選擇進行說明，然而，若配合上限之文字行具有區域座 標表，即實施跨越複數文字行之選擇。 其次，參照第9圖、第1 〇圖、第1 3圖、第1 7圖、以及 第18圖，針對以解決第5課題爲目的之第7實施形態進行說 明。和第6實施形態相同，係如「…©文法知識奁使9… 」所示，以含有由複數構詞所構成之漢字列時爲例，且假 # 設游標位於「文法」之部位時。又，和第6實施形態相同 ，只針對候補語抽出（S 1 004)、候補語顯示（S 1 005)、以及 候補語選擇（S 1 006)進行說明。 候補語抽出（S 1 004)時，係以和第5實施例所示相同之 處理利用構詞分析來產生候補語，並產生對應其之第1 3圖 所示之矩形座標表1 3 00。其次，利用表1 3 00產生分別對應 構詞及其複合語之矩形座標表1 700(第17圖）。假設表係利 用左上點座標來進行分類。 · 候補語顯示（S 1 005 )時如第18圖所示，最初係顯示含 有游標中心座標之矩形區域之構詞之矩形（1800)。其次， 按下游標按鍵1 7 6，每次都會顯示分類之表1 7 0 0之矩形。 使用者在顯示欲檢索之矩形時點按下選擇按鍵1 75，可執 行電子辭典之檢索（S 1 006)。 其次，參照第9圖、第1〇圖、第13圖、以及第19圖， 針對以解決第5課題爲目的之第8實施形態進行說明。和第 6實施形態相同，係如「…Φ文法知識全使9…」所示， -18- 1294100 (15) 以含有由複數構詞所構成之漢字列時爲例，且假設游標位 於「文法」之部位時。又，和第6實施形態相同，只針對 候補語抽出（S 1 004)、候補語顯示（S 1 00 5 )、以及候補語選 擇（S 1 006)進行說明。 候補語抽出1 004時，係以和第5實施例所示相同之處 理利用構詞分析來產生候補語，並產生對應其之第1 3圖所 示之矩形座標表1 3 00。其次，以如第19圖之1 900之形式顯 示含有游標中心座標之候補語矩形（ 1 005 )。 此處，欲選擇「文法知識」之複合語時，按下用以表 示指定選擇範圍之始點之意思之「1」之按鍵後，按下游 標移動按鍵176即可顯示如1901之矩形。再按下游標移動 按鍵176即會如1 902所示，放大選擇區域。使用者在利用 游標按鍵顯示適當之檢索對象語之矩形之時點，按下選擇 按鍵175，即可選擇檢索對象語（S1006)。 其次，參照第9圖、第10圖、第20圖、以及第21圖， 針對以解決第5課題爲目的之第9實施形態進行說明。係如 「…®文法知識處理全…」所示，以含有由3個以上之構 詞所構成之漢字列時爲例，且假設游標位於「知識」之部 位時。 由3個以上之構詞所構成之漢字列時，使用者通常會 想要以漢字列整體做爲檢索對象、或只想檢索含有游標中 心座標之構詞。因此，以下針對減輕此選擇之負擔之處理 進行說明。又，和第6實施形態相同，只針對候補語抽出 (S 1 004)、候補語顯示（S 1 005 )、以及候補語選擇（S 1 006)進 1294100 (16) 行說明。 候補語抽出（S 1 004)時，係以和第5實施例所示相同之 處理利用構詞分析來產生候補語，並產生儲存對應候補語 之文字種類及矩形座標之第20圖所示之表2000。此處之文 字種類係以候補語若爲「漢字」則屬於「漢字」、若爲「 平假名」則屬於「平假名」之方式實施各文字之種類之分 類結果。其次，會統合和含有游標中心座標之候補語（構 詞）相同文字種類之候補語之矩形，並以如第2 1圖之2 1 00 之形式顯示（S1005)。 此處，並非想要漢字列整體而係只想選擇「知識」之 候補語時，按下代表切換選擇模式之「#」之按鍵，顯示 如2 1 0 1之矩形。其次，欲選擇「知識処理」之文字列時， 按下代表指定選擇範圍之始點之「1」之按鍵後，按下游 標移動按鍵176顯示如21 02之矩形。使用者在利用按鍵操 作顯示適當之檢索對象語之矩形之時點，按下選擇按鍵 175，即可選擇檢索對象語（S 1 006)。 上述實施例時，文字之種類係只以「漢字」及「平假 名」爲例，然而，亦可應用其他如「片假名」、「英文」 、「數字」、「記號」、「外國語」等用以實施文字分類 之任意種類。此外，亦可以該構詞所屬之詞類來取代文字 之種類。 又，上述實施例時，因係以指定選擇範圍之始點、或 切換選擇之模式爲目的而按下「1」、「#」，然而，只要 爲指定著相同意思之按鍵，可指定成任意按鍵。 -20- 1294100 (17) 此外，本實施例時，係將十字記號之游標之中心座標 Λ 當做選擇資訊使用，然而，只要爲具有同樣效果之方法， · 亦可以爲其他資訊。例如，亦可以將兩個括弧顯示於顯示 器，並使用顯示之括弧之中心座標、或兩括弧之矩形座標 。此外，日文之選擇時係針對橫書進行說明，然而，亦同 樣可應用於直書。 又，上述實施例可分別實現，亦可以整體或選擇一部 份之方式來實現。 φ 【圖式簡單說明】 第1圖係本發明之實施例之攜帶型資訊終端之方塊圖 〇 第2圖係本發明之實施例之流程圖。 第3圖係本發明之實施例之流程圖。 第4圖係本發明之實施例之角度補正之說明圖。 第5圖係本發明之實施例之流程圖。 脅 第6圖係本發明之實施例之畫面。 第7圖係本發明之實施例之流程圖。 第8圖係本發明之實施例之文字列外接矩形產生方式 之說明圖。 第9圖係本發明之實施例之攜帶型資訊終端之方塊圖 〇 第1 〇圖係本發明之實施例之流程圖。 第1 1圖係本發明之實施例之矩形座標表之說明圖。 -21 - 1294100 (18) 第1 2圖係本發明之實施例之檢索候補語之選擇方式之 ~ 說明圖。 · 第1 3圖係本發明之實施例之矩形座標表之說明圖。 第1 4圖係本發明之實施例之選擇區域表之說明圖。 第1 5圖係本發明之實施例之矩形座標之說明圖。 第1 6圖係本發明之實施例之說明圖。 第1 7圖係本發明之實施例之矩形座標表之說明圖。 第1 8圖係本發明之實施例之說明圖。 鲁 第1 9圖係本發明之實施例之說明圖。 第2 0圖係本發明之實施例之矩形座標表之說明圖。 第2 1圖係本發明之實施例之說明圖。 【主要元件符號說明】 100 攜帶型資訊終端本體 110 影像輸入裝置1294100 (1) Description of the Invention [Technical Field] The present invention relates to a processing object selection method for optical character recognition of a portable information terminal. [Prior Art] A portable information terminal equipped with an image input function, such as a video camera, is actively developing a technique for performing optical character recognition on an image taken. However, the image taken by the camera of the portable information terminal tends to be inclined with respect to the screen due to hand shaking or the like. Therefore, when the character recognition result is incorrect (misread), the user must adjust the angle at the time of ingestion to re-acquire, or use an input device such as a keyboard to correct the recognition result. As described in Patent Document 1, the technique for correcting the inclination of the character string to be recognized before the identification process is performed is to detect two or more marks previously written in the paper in which the character string to be recognized is described, and to obtain the tilt obtained in this manner. The rotation is automatically performed and then recognized. Further, the English word recognition system of the portable information terminal of Non-Patent Document 1 is combined with the text recognition of the English-Japanese dictionary and the method of dictionary search. [Patent Document 1] Japanese Unexamined Patent Publication No. Hei No. 25 0 1 79 (4 to 7 and 3) [Non-Patent Document 1] H. Fujisawa, H. Sako, Y. Okada, and SW. Lee, u Information Capturing Camera and Developmental Issues, ''In Proc. Int. Conf. Document Analysis and Recognition, ICDAR799, Bangalore, India, Sep. 20-22, 1999, 1294100 (2) pp. 205-208 0 · Implementing Good Accuracy In the character recognition processing, it is extremely important that the character string of the object to be recognized in the image is not tilted. However, when a digital video camera such as a digital camera or a portable terminal with a camera is used to capture a digital image, the information terminal device is usually hand-held, so that the character of the identification object is not tilted in the image, and it is necessary to be particularly ingested. Note (First Question) 〇 For the optical type and character recognition function of images taken with a portable terminal with a camera function, when the character string is tilted, the conventional technology cannot perform character recognition or cannot have sufficient recognition accuracy. The problem. Therefore, there is a problem that the text of the identification object must be re-acquired in order to obtain the correct text recognition result. Patent Document 1 describes that the content of the tilt is detected by two or more marks described on the paper. However, this technique must print or record the mark on the paper to be recognized in advance, and is not practical when identifying a business card or the like. Further, when the character recognition processing is performed on the image taken in the past, if the image is not photographed with special attention as described above, the recognition target character string may be tilted in the image. At this time, it is necessary to perform correction of the image angle (second problem) by using the image processing software or the like before performing the character recognition processing. In addition, when there are two types of characters, such as a straight book and a horizontal book, even if the image angle is adjusted to extract the character string to be recognized, when the character recognition is performed, it is necessary to set the character to be recognized as a straight book or a horizontal book. When an image of a business card or the like is obtained and a character recognition process is performed, since there is already a format of a straight book or a horizontal book, there is a problem that the identification mode must be set each time. In addition, when newspapers, miscellaneous -6- 1294100 (3), and the like, there are a mixture of straight text and horizontal text, the direction of the text > word is a big burden (third subject). • When using a device such as a portable information terminal with a camera, it is often used in an environment where it is impossible to capture images in a stable position when the vehicle is moving. However, in the conventional technique, if the tilt of the recognized character string with respect to the image is not selected to be suppressed to a small state, accurate character recognition cannot be obtained. Therefore, the user must pay special attention to the angle adjustment (fourth problem) when taking in the image of the recognition target. In addition, when the dictionary search and the Internet search are performed using the character recognition result of the portable information terminal with the camera, the English word is distinguished by a blank word, so that the word selection of the search object is relatively easy. However, when the same processing is applied to Japanese, it is different from the English single word. Because there is no layout information using the blank space between the words to separate the text, the user must select the character string of the search object for each character, so the designation will be A considerable burden is incurred (fifth subject). SUMMARY OF THE INVENTION In view of the above problems, it is an object of the present invention to provide a system and method for reducing the burden on an ingestor when a portable information terminal using a camera capable of capturing a character recognition object, and reducing the ingestion The image is corrected to fit the angle of text recognition. Further, the object of the present invention is to provide a system and method for reducing the burden on a user to specify a character string of a search target when performing Japanese character recognition and performing dictionary and internet search. 1294100 ' (4) In order to solve the above-mentioned first problem, the angle indicator in the image of the recognition target character string can be displayed on the screen of the information terminal device at the moment. When the user observes the information displayed by the angle indicator, the mobile information terminal device or the object to be recognized is photographed at an appropriate position, and an image suitable for character recognition processing can be taken. In order to solve the above second problem, there is provided a function for rotating an already taken image and simultaneously extracting a line of the recognized object character string. The user rotates on the screen of the information terminal device by using a simple button operation or the like. • The image has been taken, and the character list candidate for the line drawing is immediately executed. At the time when the desired character string of the recognition object is extracted, the user stops the rotation processing, and when the plurality of lines are extracted, the character string is selected and the recognition processing is performed, which simplifies the character recognition processing for the image taken in the past. In order to solve the above third problem, it is provided to automatically determine whether the identification target text string is a straight book or a horizontal book. In the automatic determination of the direction of the character string of the recognition target, the aspect ratio of the rectangle is extracted by the extracted character string. Specifically, the ratio of the height and the width of the rectangle to the recognized character string is compared with the predetermined 値, and the straight text string or the horizontal character string is determined, and the identification processing is performed. When the screen orientation of the portable terminal device used is variable, the switching between the straight book and the horizontal book corresponding to the screen direction can be automatically performed. To solve the above fourth problem, the angle of inclination according to the character string of the identification object is automatically generated. Select the means to identify the object text column and then connect the rectangle. Specifically, for example, the method of detecting the tilt angle of the character string of the identification object is -8 - 1294100 (5) degrees by the method of Japanese Patent Application Laid-Open No. Hei 7- 1 4 1 465 "Tip Detection Method of Document Image", and the character string of the identification object is written as a horizontal book. Create an image that rotates the image toward a horizontal position, and creates an image that rotates the image toward a vertical position. Then, <, the rectangle of the recognized object character string is generated, and the rotated circumscribed rectangle image added to the rotated image is created. Then, the rotated circumscribed rectangular image is rotated to return to the tilt angle of the reference character string, and displayed on the display device of the information terminal device. In order to solve the above fifth problem, a means is provided for performing word formation analysis on the Japanese character recognition result, and automatically generating a character string candidate for the search target, and the user selects each candidate or selects a combination of each character string candidate. Displaying the tilted image of the image of the recognized object in a manner that is easy to visually understand allows the user to know the tilt state of the image, and it is easy to take the image at an angle suitable for the character recognition processing. Moreover, in the image that has been taken in the past, the oblique image is directly edited, and the character recognition processing can be performed, so that it is not necessary to take another image. Moreover, since it is difficult to correct the angle, the character string can be selected in an inclined state. Therefore, the character recognition processing can be performed on the image having a certain degree of inclination. [Embodiment] Referring to FIGS. 1 to 20, a preferred embodiment of the present invention is described. Be explained. The text recognition method of the present invention can be applied to a mobile information system for reading and recognizing a business card, etc., for example, a text recognition machine that can also be applied to a mobile phone. FIG. 1 is a block diagram of a portable information terminal to which the present invention is applied. -9 - 1294100 (6) Example. In this embodiment, the portable information terminal body 100; the image input device 1 10 for the camera and the scanner for optically inputting the identification card, etc.; for displaying the image of the identification object, the character recognition result, and A display device 120 such as a CRT or a liquid crystal such as a cursor 121; an input device 1 30 that is operated by a user operating button 1 31; and a terminal device 1 for carrying out overall control of the portable information terminal; The control unit 140 and the character line extracting unit 151 in the frame, the character recognition unit 150 for performing the character recognition processing 152, and the image processing unit for the quantization function 161 and the image rotation processing function 162 having the character line inclination 160. The character recognition unit 150 and the image processing unit 160 may be software functions, and may operate on the same circuit as the control unit 140. The input device may be a general device such as a button. However, in order to improve operability, the display device 120 and the input device 130 may be realized by a display device such as a touch panel and an input device. Fig. 2 is a diagram showing an example of a flowchart in the process of character recognition processing for the purpose of solving the first problem (first embodiment). When the user starts the OCR function, the moving image input from the image input device 1 1 is displayed on the display device 1 2 0 (S 2 0 1). The character recognition unit 150 immediately performs a character line extraction process (S2 02) for the area where the cursor 121 of the cross type or the like displayed on the image display device 120 is located, and displays an circumscribed rectangle for extracting the extracted character line (S203). At the same time, the image processing unit 1 60 performs quantization of the inclination of the character string ‘the quantized 显示 is displayed on the angle indicator 1 23 in the form of a bar graph or the like (S204). After quantification, as long as it can reflect the inclination of the text column and image 1294100 (7), for example, it is proportional to the reciprocal of the angle formed by the line of characters and the edge of the image, and the angle indicator is small when the tilt is small. 1 2 3 will show a larger flaw, and the user can use the intuition to perform the operation. When the user presses the ingest button (S20 5), the processing returns to the step (S201), and the processing (S201 to S204) is repeated to continue the display of the update screen. The user refers to the angle indicator 1 23 and moves the terminal device or the recognized object, and presses the ingest button at the appropriate ingest position (S20 5) to perform image pickup (S206). Next, when the identification button is pressed (S207), the character recognition processing 210 is performed on the character string in the outer rectangle displayed in the step (S203), and the recognition result is displayed (S211). When the capture button (S20 5 ) is pressed, if the cancel button is pressed (S208), the captured image is discarded and the process returns to the step (S201). When the other button is pressed, the necessary action is performed (S209). Fig. 3 is a diagram showing an example of a flowchart in the process of character recognition processing for the purpose of solving the second problem (second embodiment). The user of the portable information terminal performs loading of the previously taken still image (S301), and the still image is displayed on the display device 120 (S308). The character recognition unit 150 immediately executes the character line extraction (S 3 03 ), the circumscribed rectangle display (S 3 04), and the angle indicator display (S 3 0 5 ). At this point, processing proceeds to the stage of waiting for the user's key input (S 3 06). When the user presses the recognition execution button (S 3 07), the character recognition processing is immediately performed on the character string displayed inside the rectangle of the step (S 3 04) (S3 10), and the character recognition result is displayed (S3 11). . When the user presses the arrow and presses the 1294100 (8) button (s 3 0 8 ), the button is pressed to rotate the image to the right or left by a certain angle (S3 09). At this time, the center of rotation is the center of the character line rectangle, but the rotation processing can be performed centering on one of the points on the image designated by the user. When you hold down the arrow button, the image rotates continuously and the circumscribing rectangle display is updated as you see it. In consideration of the convenience of the user, it is also possible to stop the action of pressing the arrow button (the finger leaves the button) instead of the operation of recognizing the execution button. Fig. 4 is a view showing an example of an image displayed on the display unit 120 in the first embodiment and the second embodiment. In the first embodiment, the display unit 120 displays the animation input by the video input device 110. In the second embodiment, the display unit 1120 displays the previously captured still picture 400. At this point, the recognition target character string 4 0 1 in the image 400 is tilted. The center of the screen shows the cross cursor of the center of rotation 4 0 2 . The character recognition unit 150 generates a rectangle 403 for framing the recognition target character string 401 and displays it on the screen. The angle indicator displays a bar graph (404) indicating the degree of tilt of the recognized object string 401. In the first embodiment, the image displayed on the screen is rotated by a portable handheld information terminal (camera) (4〇6), and in the second embodiment, the arrow key (405) is operated to rotate. The image displayed on the screen. The tilt of the recognized object character string 40 1 becomes smaller as the image is rotated, and the circumscribed rectangle also produces a synchronized shape change (407). In addition, the angle indicator uses a bar graph to display a larger 値 (408) representing a smaller slant of the identification object text -12 - 1294100 Ο) column 401. The user repeatedly performs the rotation operation of the image, and performs character recognition processing at a position where the angle indicator is displayed at a larger position, thereby obtaining a highly accurate character recognition result. Fig. 5 is an example of a flowchart (third embodiment) in the case of performing text recognition processing for the purpose of solving the third problem. Since the steps (S501 to S506) are the same as the steps (S201 to S209), the description thereof will be omitted. When the identification button is pressed (S 5 04), the aspect ratio (ratio of height and width) of the rectangle to be recognized in the character string of the recognition target is calculated, and compared with the predetermined 値α (S5 07). When the aspect ratio is larger than the predetermined value 値α, it is determined that the straight book character string 'the parameter setting for recognizing the straight character character string is set (S 510), the character recognition processing is executed (S51 1), and the result is displayed (S512). Similarly, if the aspect ratio is not larger than the predetermined 値α, then the aspect ratio is compared with the specified 値/3 (S 5 08). When the aspect ratio is smaller than the predetermined 値Θ, it is judged as a horizontal character string, and parameter setting for recognizing the horizontal character string is performed (S5 09), and character recognition processing (S51 1) is executed and the result is displayed (S512). When the aspect ratio is α or less and Θ or more, it is considered that the angle correction is not performed on the character string and the recognition processing is not performed. Fig. 6 is a view showing an example of an image displayed on the display unit 1 in the third embodiment. When the character string 601 of the horizontal book is 601, the height 602 of the circumscribed rectangle is smaller than the width 603. If the height of the circumscribed rectangle is the width of the circumscribed rectangle to define the aspect ratio, and the aspect ratio is less than 1, the circumscribed rectangle is horizontally long. For example, when the specification 値Θ for comparison with the aspect ratio is set to 〇 5 , when the aspect ratio is lower than 〇 5 , the character recognition is started, and the necessary setting of the horizontal character string is automatically implemented. Similarly, when the straight text column 604, the height 605 of the 1294100 (10) circumscribed rectangle is greater than the width 606. If it is the same as the above definition of the aspect ratio, when the aspect ratio is greater than 1, the circumscribed rectangle is vertically long. For example, when the specification 値α for comparison with the aspect ratio is set to 1-5, the character recognition is started in the state where the aspect ratio is higher than 丨.5, and the necessary setting of the straight text column is automatically performed. Fig. 7 is an example of a flowchart (fourth embodiment) at the time of performing the character recognition processing of the means for solving the fourth problem. When the character recognition target image is input from the image input device 110 (S 7 0 1), the image processing unit 1 60 immediately calculates the angle of the recognition target character string with respect to the image (S 702), and performs the input on the character recognition target image. Angle rotation correction. In the correction direction, the horizontal character column is rotated in the horizontal direction of the screen, and the straight text column is rotated in the vertical direction of the screen. Next, the character string is extracted for the rotated image, and a circumscribed rectangle is added to the rotated image (S 704), and the image is stored in the image buffer. The image stored in the image buffer can be the whole image, or it can be attached to the inside of the rectangle. Then, the reverse rotation of the image after the rotation is performed by the angle portion detected in the step (S702), and an image which is inclined as the original input image is created and displayed on the display portion 1208 of the terminal device. When the user performs any operation, the process returns to the step (S701), and the processing of the steps (S701 to S707) is repeatedly performed for the new input image. When the identification button is pressed, the image stored in the image buffer in step (S708) is read (S710), the character recognition processing is performed on the image (S71 1), and the character recognition result is displayed (S 7 1 2 ) . 1294100 (11) Fig. 8 is an example of an image state processed by the fourth embodiment. The recognition target character string 082 on the identification target image 801 input from the image input device 1 10 is inclined to the screen. The image processing unit 〇6 detects the angle 8〇3 of the recognized object character string 802 and the edge of the screen, and corrects the recognition target character string by the image rotation angle 803 ′ to form a rotation correction image 804. The character recognition unit 150 performs character line extraction on the recognition target character string 806 on the rotation correction image 804, and adds a circumscribed rectangle 806. The image processing unit 160 performs the reverse rotation of the angle 803 on the rotation corrected image 8〇4 to generate an image 807 which is returned to the same angle as the original recognition target image 810, and is displayed on the display device 120. The above first to fourth embodiments may be implemented separately, or all or part of them may be selected. Next, a fifth embodiment for solving the fifth problem will be described with reference to Figs. 9 to 12 . Fig. 9 is an example of a block diagram of a portable information terminal to which the fifth embodiment is applied. The difference between the first and the first figures is the addition of the search term extracting unit 170 and the electronic dictionary 171. The first diagram is an example of a flowchart in the case where the apparatus of Fig. 9 implements character recognition and dictionary search processing by means of solving the fifth problem. The character recognition target image is input from the image input device 1 10 (S1001), and after the image processing unit 160 performs appropriate image processing on the source image, the character line extracting unit 1 51 of the character recognition unit 150 extracts the character line (S 1 0 0 2 ). Thereafter, the character line recognition unit 1 52 performs character cutting on the extracted character lines and outputs the recognition result of each character (S 1 003 ). The recognition result includes the text code of each character and the corresponding 1294100 (12) rectangular coordinates. The result of the character recognition is input to the search term candidate extracting unit 1 7 0 ', and the continuous character string is decomposed into the word formation such as a single word by using the word formation analysis (S 1 〇 0 4). For example, when the character string of the "grammar Φ knowledge $•使? &" displayed on the display of the portable information terminal of Fig. 9 is generated, the table 1 1 00 as shown in Fig. 1 is generated. A character string that is decomposed using word formation analysis and corresponding rectangular coordinates are stored. The candidate of the search object is displayed on the display of the portable information terminal using the data of Table 1 1 S ( (S 1 005). For example, compare the center coordinates of the cursor with the candidate rectangular coordinates of the table 1 1 , and display the rectangular coordinates of the candidate containing the center coordinates of the cursor on the display in the form shown as 1 200 in Fig. 12 . Next, by pressing the downstream label moving button 174 or 176, the rectangular display of the search target language can be moved as shown in 1201, and when the rectangular table to be searched is displayed, the selection button 175 is pressed to determine the search target language (S1006). The determined search target language is output to the electronic dictionary section 1 71. The electronic dictionary section 1 7 1 treats the search target language as a keyword search (S1 〇〇 7) of the electronic dictionary and displays the search result on the display (S 1 00 8). In the present embodiment, the word formation is used to generate a candidate for the search target. However, it is also possible to separate and generate an alternate when the type of the character is changed by converting the "kanji" in the character string of the recognition result to the "hiragana". Language approach. Alternatively, it may be judged by using geometric information such as the type of the character, the blank, and the position where the character size is changed. Next, referring to Fig. 9, Fig. 10, and Fig. 13 to Fig. 15, -16-1294100 (13) A sixth embodiment for the purpose of solving the fifth problem will be described. In this case, the example is as follows: "... Φ grammar knowledge"), as shown in the example of a Chinese character column consisting of * plural words, and the cursor is assumed to be in the text of "grammar knowledge" One of the columns. The user may only use the wording of "grammar" or "knowledge" as a search object or the entire compound of "grammar knowledge" as a search target. Therefore, the following describes the process of reducing the burden of this selection. Since the part different from the processing flow of the sixth embodiment is a candidate word extraction (S 1 004), a candidate display (S 1 005 ), and a candidate selection (S 1 006), it is only for The foregoing processing will be described. When the candidate is extracted (S 1 00 4), the candidate word is generated by the word formation analysis in the same processing as that shown in the fifth embodiment, and the rectangular coordinate table 1 3 00 corresponding to the first figure is generated. . Next, an area coordinate table 1 400 for selecting "grammar", "grammatical knowledge", and "knowledge" from Table 1300 is generated. This table is a table for the purpose of displaying the rectangle corresponding to its candidate when the cursor center coordinates enter the area. ® Figure 15 is a schematic diagram of the X coordinate of the rectangular coordinates of the table. The selection areas corresponding to "grammar", "grammar knowledge", and "knowledge" are 1 500, 1 5 0 1 , and 1 5 0 2, respectively, according to the center coordinates of the cursor located in the area - from Table 1 400 Select the rectangular coordinates used to display the search candidate. The selected rectangular coordinates are displayed on the display (S 1〇〇5). Figure 16 shows an example of a rectangle. (a) shows an example of the display when the cursor is in the selection area of "Knowledge", and (b) and (c) are examples of the display when the cursor is located in the selection area of "Grammar Knowledge" and "Literature". When the user displays the rectangle of the candidate to be searched in -17- 1294100 (14), press the selection button 1 7 5 to select the search target (S 1 006). This embodiment describes the character_column selection in one character line. However, if the character line matching the upper limit has an area coordinate table, the selection across the plural character lines is implemented. Next, a seventh embodiment for solving the fifth problem will be described with reference to Fig. 9, the first drawing, the third drawing, the seventeenth, and the eighteenth. The same as the sixth embodiment, as shown in "...© grammar knowledge 99...", for example, when a Chinese character column composed of plural word formations is included, and the cursor is located at a portion of the "grammar". Further, as in the sixth embodiment, only the candidate extraction (S 1 004), the candidate display (S 1 005), and the candidate selection (S 1 006) will be described. When the candidate is extracted (S 1 004), the candidate word is generated by the word formation analysis in the same processing as that shown in the fifth embodiment, and the rectangular coordinate table 1 3 00 shown in Fig. 13 corresponding thereto is generated. Next, a rectangular coordinate table 1 700 (Fig. 17) corresponding to the word formation and its compound word is generated by using Table 1 3 00. Assume that the table is classified using the top left coordinate. • When the candidate display (S 1 005 ) is as shown in Fig. 18, the rectangle (1800) of the word formation of the rectangular area containing the coordinates of the center of the cursor is initially displayed. Secondly, press the downstream button 1 7 6 to display the rectangle of the classification table 1 7 0 0 each time. When the user presses the selection button 1 75 while displaying the rectangle to be searched, the electronic dictionary search can be performed (S 1 006). Next, an eighth embodiment for solving the fifth problem will be described with reference to FIG. 9, FIG. 1, FIG. 13, and FIG. As in the sixth embodiment, as shown in "... Φ grammar knowledge is 9...", -18- 1294100 (15) is taken as an example of a Chinese character column composed of plural word formations, and the cursor is assumed to be "grammar When the part is. Further, similarly to the sixth embodiment, only the candidate extraction (S 1 004), the candidate display (S 1 00 5 ), and the candidate selection (S 1 006) will be described. When the candidate is extracted 1 004, the wording analysis is used to generate the candidate in the same manner as in the fifth embodiment, and the rectangular coordinate table 1 3 00 shown in Fig. 3 corresponding thereto is generated. Next, the candidate rectangle (1 005 ) containing the center coordinates of the cursor is displayed in the form of 900 as shown in Fig. 19. Here, when you want to select the grammar of "grammar knowledge", press the button for "1" indicating the start point of the specified selection range, and press the downstream button 176 to display the rectangle such as 1901. Pressing the downstream label movement button 176 will zoom in on the selection area as indicated by 1 902. When the user displays the rectangle of the appropriate search target word by using the cursor button, the user selects the button 175 to select the search target language (S1006). Next, a ninth embodiment for solving the fifth problem will be described with reference to FIG. 9, FIG. 10, FIG. 20, and FIG. For example, as shown in "...® Grammar Knowledge Processing...", for example, a Chinese character column consisting of three or more constructive words is used, and the cursor is assumed to be in the "Knowledge" part. When a Chinese character string consisting of three or more word formations is used, the user usually wants to use the entire Chinese character column as a search target, or only wants to search for a word containing a cursor center coordinate. Therefore, the following processing for reducing the burden of this selection will be described. Further, as in the sixth embodiment, only the candidate extraction (S 1 004), the candidate display (S 1 005 ), and the candidate selection (S 1 006) are described as 1294100 (16). When the candidate is extracted (S 1 004), the candidate word is generated by the word formation analysis in the same processing as that shown in the fifth embodiment, and the character type and the rectangular coordinates of the corresponding candidate are stored as shown in Fig. 20 of the rectangular coordinates. Table 2000. The type of the text here is the classification result of the type of each character in the case where the candidate is "Chinese character" for the "Chinese character" and the "hiragana" is the "hiragana". Next, the rectangle of the candidate having the same character type as the candidate (conformation) of the cursor center coordinates is integrated and displayed in the form of 2 1 00 as shown in Fig. 2 (S1005). Here, when it is not intended to select the "Knowledge" candidate for the entire Chinese character list, the button "#" representing the switching selection mode is pressed to display a rectangle such as 2 1 0 1 . Next, when a character string of "knowledge processing" is to be selected, a button of "1" representing the start point of the specified selection range is pressed, and a rectangle such as 21 02 is displayed by the downstream label movement button 176. When the user operates the rectangle of the search target word by the button operation, the user selects the button 175 to select the search target language (S 1 006). In the above-mentioned embodiments, the types of characters are only "Chinese characters" and "hiragana". However, other such as "katakana", "English", "number", "mark", "foreign language", etc. can be applied. Any type used to implement word classification. In addition, the word class to which the word formation belongs may be substituted for the type of the word. Further, in the above-described embodiment, "1" and "#" are pressed for the purpose of specifying the start point of the selection range or switching the selection mode. However, if the button is assigned the same meaning, it can be designated as arbitrary. button. -20- 1294100 (17) In addition, in the present embodiment, the center coordinate of the cursor of the cross mark is used as the selection information. However, as long as it has the same effect, it can also be other information. For example, you can also display two brackets on the display and use the center coordinates of the brackets displayed, or the rectangular coordinates of the two brackets. In addition, the choice of Japanese is explained for Hengshu, however, it can also be applied to straight books. Moreover, the above embodiments may be implemented separately or in a whole or in part. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a portable information terminal according to an embodiment of the present invention. FIG. 2 is a flow chart of an embodiment of the present invention. Figure 3 is a flow chart of an embodiment of the present invention. Fig. 4 is an explanatory view of angle correction of an embodiment of the present invention. Figure 5 is a flow chart of an embodiment of the present invention. Figure 6 is a screen of an embodiment of the present invention. Figure 7 is a flow chart of an embodiment of the present invention. Fig. 8 is an explanatory view showing a manner of generating a circumscribed rectangle of a character string in the embodiment of the present invention. Figure 9 is a block diagram of a portable information terminal in accordance with an embodiment of the present invention. Figure 1 is a flow chart of an embodiment of the present invention. Fig. 1 is an explanatory view of a rectangular coordinate table of an embodiment of the present invention. -21 - 1294100 (18) Fig. 1 is a diagram showing the selection of search candidates in the embodiment of the present invention. Fig. 13 is an explanatory view of a rectangular coordinate table of an embodiment of the present invention. Fig. 14 is an explanatory view of a selection area table of an embodiment of the present invention. Fig. 15 is an explanatory view of a rectangular coordinate of an embodiment of the present invention. Fig. 16 is an explanatory view of an embodiment of the present invention. Fig. 17 is an explanatory view of a rectangular coordinate table of an embodiment of the present invention. Fig. 18 is an explanatory view of an embodiment of the present invention. Lu. Fig. 9 is an explanatory view of an embodiment of the present invention. Fig. 20 is an explanatory view of a rectangular coordinate table of an embodiment of the present invention. Fig. 2 is an explanatory view of an embodiment of the present invention. [Main component symbol description] 100 Portable information terminal body 110 Image input device
1 20 顯示裝置 12 1 游標 122 外接矩形 123 角度指示器 130 輸入裝置 131 按鈕 140 控制部 150 文字辨識部 151 文字行抽出 -22- 1294100 (19) 152 文 字 辨 識 處 理 160 影 像 處 理 部 161 量 化 機 能 162 影 像 旋 轉 處 理 機能 170 檢 索 語 抽 出 部 171 電 子 辭 典 174 游 標 移 動 按 鍵 175 々BB 培 擇 按 鈕 176 游 標 移 動 按 鍵 400 影 像 40 1 辨 識 對 象 文 字 列 402 十 字 游 標 403 外 接 矩 形 404 條 形 圖 405 箭 頭 鍵 406 攜 帶 型 資 訊 終 丄山 m 408 條 形 圖 60 1 文 字 列 602 局 603 寬 604 文 字 列 605 局 606 寬 80 1 辨 3哉 對 象 影 像1 20 Display device 12 1 Cursor 122 External rectangle 123 Angle indicator 130 Input device 131 Button 140 Control unit 150 Character recognition unit 151 Character line extraction -22- 1294100 (19) 152 Character recognition processing 160 Image processing unit 161 Quantization function 162 Image Rotation processing function 170 Search term extraction unit 171 Electronic dictionary 174 Cursor movement button 175 々 BB selection button 176 Cursor movement button 400 Image 40 1 Identification object character column 402 Cross cursor 403 External rectangle 404 Bar graph 405 Arrow button 406 Portable information Final Mountain M 408 Bar Chart 60 1 Character Column 602 Bureau 603 Width 604 Character Column 605 Bureau 606 Width 80 1 Discriminate 3 哉 Object Image
-23- 1294100 (20) 802 辨 803 角 804 旋 806 辨 807 影 1100 表 1300 矩 1400 區 1500 CBB m 150 1 CEB 培 1502 々BB m 1700 矩 1800 矩 2000 表 識對象文字列 度 轉補正影像 識對象文字列 像 形座標表 域座標表 擇區域 擇區域 擇區域 形座標表 形-23- 1294100 (20) 802 803 803 angle 804 rotation 806 807 shadow 1100 table 1300 moment 1400 area 1500 CBB m 150 1 CEB training 1502 々 BB m 1700 moment 1800 moment 2000 recognition object text column degree to correct positive image identification object Text column pictogram coordinate table domain coordinate table selection region selection region selection region shape coordinate table shape