查詢結果分析
來源資料
相關文獻
- Character Segmentation in Machine Printed Chinese Characters
- How to Improve Chinese Student's Written English
- 美哉漢字--傳統民間美術字
- Chinese Morphology: An Exploratory Study of Second Language Learners' Acquisition of Compounds
- The German Complex Affix: Word Formation and Foreign Language Instruction
- 相機取像的影像文字擷取與切割
- 新聞報道記事における臨時一語
- Metonymies in Taiwanese Southern Min: Body Parts and Persons
- 英語複合字之重音及其意義
- Are There Any Rules for Teaching Compounds?
頁籤選單縮合
題名 | Character Segmentation in Machine Printed Chinese Characters=中文印刷字體之文字切割 |
---|---|
作者姓名(中文) | 楊翊中; 傅楸善; 陳良華; | 書刊名 | 電腦學刊 |
卷期 | 10:1 1998.03[民87.03] |
頁次 | 頁1-14 |
分類號 | 312.74 |
關鍵詞 | 文字切割; 複合字; 重疊字; 相連字; Character segmentation; Compound character; Overlapping character; Touching character; |
語文 | 英文(English) |
中文摘要 | 文字切割是文字辨識(Optical Character Recognition)中最重要的前置處理之 一。我們的演算法能夠將影像檔中的中文印刷字切割出來,即使這些中文字在經由攝影或掃 圖機等將文件影像化的過程中,有因文字排列太緊密或機械引起的失真所產生的文字重疊及 文字相連問題。另外,中文字中包含形如'州'、'明'等由數個模組形成的文字,我們的演算 法提供一序列式的處理來解決這三種問題。 我們使用一個簡化的動態程式(dynamic programming)來決定那幾個模組該形成一個文 字;分析各筆劃的終止位置(connected component analysis)來處理文字重疊問題;結合寬高 比(aspect ratio)、由背景轉前景次數(black to white transition)、及二階微分來解決 文字相連問題。 我們的演算法並不使用辨識的結果來輔助文字切割,同時因係針對純粹的黑白而非灰階 影像,故處理速度較快。 |
英文摘要 | Character Segmentation is one of the most important preprocessings for optical character recognition. Our algorithm can extract individual machine printed Chinese characters from text image. Our algorithm uses a modified dynamic programming technique to solve compound (or merged) character segmentation problem and uses the connected component analysis to solve the overlapping character segmentation problem. We combine the aspect ratio, black-to-white transition, distances between black pixels at segmentation position candidates with the same number of black-to-white transition, and the second derivative of vertical projection to solve touching character segmentation problem without recognition-based method. Those segmentation problems are solved in sequence. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。