基于多阈值归一化分割的模糊图像边缘分割算法

Segmentation Algorithms of Fuzzy Image Edge Based on Multi-threshold Normalized Segmentation

摘要:

模糊图像边缘的像素特征较为复杂,一般需要采用多个阈值作为分隔约束条件的方法来进行图像边缘分割,但是该方法存在诸如多阈值无法形成统一标准、边缘提取过程需要多次校对,以及效率较低等缺点。提出一种基于多阈值归一化分割的模糊图像边缘分割算法,通过设计超像素网格对模糊图像边缘特征的像素进行匹配,分析模糊图像的反调张量信息,并根据不同张量信息对多阈值进行归一化,以及采用灰度窗口相关系数匹配方法,将获得的多阈值归一化结果分别覆盖图中的单一目标对象,以实现模糊图像的边缘分割。实验表明,利用该算法进行模糊图像边缘分割能较好地获取图像的边缘细节特征,使得边缘具有更好的连线段连通性和宽度一致性。?更多还原

Abstract:

The pixel features of edges of the blurred image are much complex, thus it generally uses multiple threshold as space constraints, but such problems exist in this method as it cannot form a unified standard threshold, edge detection process needs to be checked for many times, and the efficiency is low. In this paper, put forward is a more normalized segmentation algorithm of fuzzy image edge based on multi-threshold normalized segmentation. For the new segmentation algorithm, superpixel grid is designed to make pixel matching of the fuzzy image edge, tensor information of the fuzzy iamge is analyzed, and according to different tensor information, normalizations are performed on multiple thresholds. And with the gray window correlation matching method, the obtained multi-threshold normalization respectively overlays of the single target, thus realizing edge segmentation of blurred images. Experiments show that the proposed algorithm for fuzzy image edge segmentation can well reflect the image edge detail features, making the edge present better connectivity and width uniformity.