文字画像の修復による低解像度印刷漢字の認識
スポンサーリンク
概要
- 論文の詳細を見る
A restoration method is presented for low-resolution printed kanji character recognition. First, character images are converted into binary character patterns after expanding and blurring. Next, the binary character patterns are modified by the system using ridge point and ravine one. The feature vectors of the modified character patterns which include the information of spatial structure are extracted by phrase feature distribution method and they are fed in an classifier. It is verified by experiments using the database ETL2 that proposed method improved recognition accuracy from 89.1% to 99.4%.
- 長崎大学の論文