In this paper, the embedment of class information into each character image is investigated for camera-based character recognition as easy and accurate as bar-code reading. Each character image is printed with a horizontal stripe pattern, called a cross ratio pattern, and the class information is represented as a cross ratio derived from the pattern. Since the cross ratio is invariant to projective distortion, the class information is extracted correctly regardless of camera angle. Experimental results showed that the cross ratios are extracted from distorted character images with very high accuracy and thus very effective to attain high character recognition rates.
|出版物ステータス||出版済み - 12 1 2005|
|イベント||1st International Workshop on Camera-Based Document Analysis and Recognition, CBDAR 2005 - Seoul, 大韓民国|
継続期間: 8 29 2005 → 8 29 2005
|その他||1st International Workshop on Camera-Based Document Analysis and Recognition, CBDAR 2005|
|期間||8/29/05 → 8/29/05|
All Science Journal Classification (ASJC) codes
- Computer Vision and Pattern Recognition