基于同质性和TASOM网络的边缘检测算法

本文提出了一种新的基于同质性和时间自适应自组织神经网络（TASOM）的边缘检测算法。算法首先利用同质性检测到图像的候选边缘点，并根据这些点构造TASOM网络结构；其次，利用Canny算子检测到的边缘点作为网络训练的特征点；最后，通过比较控制点和特征点调整TASOM网络的连接权，实现对网络神经元的自动添加和删减，达到保持边缘的连续性和删除冗余边缘点的目的。试验结果表明所提算法优于传统的边缘检测方法。
关键词同质性；Canny算子；TASOM；边缘检测；特征点
Abstract This paper presents a new algorithm for edge detection based on homogeneity and time adaptive self-organizing neural networks (TASOM). Firstly, the method uses homogeneity to extract the candidate edge points, and constructs the TASOM network according to these points; and then, make the edge points detected by Canny operator as
feature points of the TASOM network; finally, the weights of TASOM are updated to achieve the adding and deletion of neurons automatically by comparing the difference between control points and feature points; consequently, the aims that remaining the continuity of the edges and deleting the redundant points are achieved. The experimental results show that the
proposed method is superior to other algorithms.
Keywords homogeneous；canny operator；TASOM；edge detection；feature points