基于凹凸特性的非限制粘连手写数字串切分

针对现有的切分算法结构复杂,时间和空间复杂度高等不足,提出了一种基于凹凸特性的非限制粘连手写数字串切分的新方法。首先计算数字串图像的赋值背景,然后从中提取凹凸特性,找到切分区域,最后在切分区域内提取切分线。该方法简单快速,在提高切分正确率的同时也降低了复杂度。利用NISTSD19收集到的样本进行实验,正确率高达97.5%,切分时间也大大缩短。

A New Segmentation Method of Unconstrained Handwritten Connected Numeral String Based on Concave-and-Convex Feature

Considering such defects of the present segmentation methods as complex structure and so on,a new method based on con-cave-and-convex feature is proposed for unconstrained handwritten connected numeral strings segmentation.First value_associated background of string's image is found,convex-and-concave feature is then extracted.At last,the convex areas for segmentation are found.This method is simple and fast,which reduce the time and space complexity.Test results on collection of samples from NIST SD19 show that the system achieve an accuracy of 97.5%,and the time also is shorten greatly.