查詢結果分析
來源資料
頁籤選單縮合
題名 | 光學中文多種印刷字體自動辨認個人電腦系統 |
---|---|
作者姓名(中文) | 張耿豪; | 書刊名 | 電信研究 |
卷期 | 20:1 1990.03[民79.03] |
頁次 | 頁59-75 |
專輯 | 光學中文文字識別專輯 |
分類號 | 312.74 |
關鍵詞 | 中文; 光學; 印刷; 字體; 電腦; 辨認; |
語文 | 中文(Chinese) |
中文摘要 | 本文介紹一套多種字體、不同大小的中文印刷字辨認系統。我們採用的辦認待徵是以機種方向(水平、垂直、45°、135°)筆劃之相關位置與背景之關係,經過壓縮,正規化的處理,然後合併成一組具有可加成性的辦認特徵符號值。經由失真距離函數,算出每一個字的失真距離,其失真距離最小者即為辨認的結果。對於不同字體(明體、黑體、楷書)、不同大小(80×80、64×64、48×48)各5401備字,測試結果辦認率在98%以上。 |
英文摘要 | The paper describes a newly developed multi-font multi-size printed Chinese character recognition system. The multi-font character features are extracted from the relations between character background and relative location among four-direction strokes (i.e., horizontal, vertical, 45° and 135°). The multi-font features are then converted to one of the several symbols through a compression process and then a normalization process. The final feature in the recognition phase is obtained by counting each symbol separately. In the recognition phase, the distortion is measured by a distortion distance function. A test character is identified as the reference character with the smallest distortion measure. Experiments have been carried out on the data of multi-font character with regular script, Ming style and Boldface style with 80×80, 46×46 and 48×48 resolutions. Each test is carried out with 5401 Chinese characters and results in 98% recognition rate. A higher recognition rate has been achieved as an algorithm to distinguish similar characters is attached. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。