Chinese Character Stroke Based Orientation Identification for Scanned Documents

Scanned document image orientation identification is a common computer vision task, especially in research on document analysis and recognition. Most of the existing documents image orientation identification methods are focus on English scripts. A novel stroke based orientation identification method is presented for Chinese scanned documents. Two typical Chinese character stroke features, which are named as roof cross and corner cross, are extracted to determine the orientation of documents. Compared with previous methods, proposed method is suitable for handling Chinese document images. The experimental results demonstrate the feasibility and effectiveness of proposed method in 318 Chinese and English document images.