基于两方向动态时间规整的无分割手写汉字检测

中文文本布局复杂,汉字种类多,书写随意性大,因而手写汉字检测是一个很有挑战的问题。本文提出了一种无分割的手写中文文档字符检测的方法。该方法用SIFT定位文本中候选关键点,然后基于关键点位置和待查询汉字大小来确定候选字符的位置,最后用两个方向动态时间规整(Dynamic Time Warping, DTW)算法来筛选候选字符。实验结果表明,该方法能够在无需将文本分割为字符的情况下准确找到待查询的汉字,并且优于传统的基于DTW字符检测方法。

Two-directional dynamic time warping based Chinese handwritten segmentation-free word spotting

Large variety of Chinese characters and handwriting styles and the complexity of Chinese handwritten documents lead a huge challenging for the Chinese handwriting word spotting. A segmentation-free word spotting method for Chinese handwritten documents is proposed in this paper. Firstly, we use the SIFT keypoint detector to detect the candidate keypoints in document images. Then, the candidate character regions are determined by the locations of keypoints and the size of query word image. At last, the two-directional Dynamic Time Warping (DTW) is applied to refine the candidate regions. The experimental results show that our method can detect the query word with a high mean average precision and the two-directional DTW outperforms the traditional DTW.