基于方向链码去除骨架图像毛刺算法

Skeleton blur removing algorithm based on direction chain code

毛刺的去除是模式识别中图像预处理的重要组成部分。针对毛刺的去除问题,提出一种基于方向链码去除毛刺的方法。首先利用节点和端点不同的八邻域特性,直接寻找节点;然后从节点开始,通过设定初始扫描方向数,可以首先扫描标记想要删除的毛刺或分支;再通过设定长度阈值,在方向链码标记跟踪图像的基础上,计算被跟踪分支的长度,并与设定的长度阈值比较;最终达到有效删除毛刺或不必要分支的目的。实验结果表明,所提方法在车牌识别和字符识别的图像预处理中,可以有效去除任意方向的毛刺,并保留图像的主体信息,对单像素骨架图像去除毛刺有一定的可行性。

Removing burr is an important part of image preprocessing in the pattern recognition. This paper put forward a method of deburring based on the direction chain code in accordance with the problem of removing burr. At first, this method utilized the difference of eight adjacent areas＇ characteristics of node and endpoint, searched for node directly, and then started from the node through the setting of initial scanning direction number, firstly scanned and marked burr or branch that need deleting, again through the setting of length threshold value, on the basis of the direction chain code labelling and tracking image, calculated the length of the tracked branch, and compared with the threshold of setting, finally effectively deleted burr or unnecessary branch. The experimental results show that this method can effectively remove the burr with any direction in the image preprocessing of license plate recognition and character recognition, and can retain the main body information of skeleton. So, the survey has some feasibility.