特征离散点计算在手写文本行分割中的应用

Using feature discrete-point computing in handwritten documents line segmentation

将图像分析实践中的经验知识与粒计算的基本思想相结合,总结形成了特征离散点计算,并将其应用于自然手写汉字文本行分割当中。在特征离散点计算的结构化问题求解框架下,提出了一种反馈式分列行投影文本行分割方法,分为特征离散点选择、特征离散点采样与优化、特征离散点编组与反馈以及行边缘优化四个环节。该方法在哈尔滨工业大学多人手写数据库上取得了相对以往算法较好的实验结果,同时分割速度较快。

Combined the empirical knowledge of image analysis with the concept of granular computing, feature discrete-point computing is summarized and developed in this paper, which is used in the line segmentation of natural handwritten Chinese documents. According to the structured problem solving framework of the feature discrete-point computing, a line segmentation method based on feedback row projection with sub-block is proposed, which are divided into four stages, including feature discrete-point’s selecting, sampling and optimizing, grouping and feedback, and line-edge optimizing. This method achieves better experimental results on Harbin Institute of Technology-Multiple Writers Database with faster segmenting speed, comparing with existing methods.