摘要
该文设计完成了一种具有实用意义的形态学开、闭滤波的神经网络模型及其滤波参数的优化训练算法。实验结果表明该方法设计简便,实用性强且易于推广,对提高形态滤波性能效果明显。分析表明,形态滤波器可分解为形态滤波运算和结构元素选择两个基本问题。形态滤波运算规则已由定义本身确定,于是形态滤波器的最终滤波性能就仅仅取决于结构元素的选择。进行自适应优化训练的目的正是使结构元素具有图像目标的形态结构特征,从而使形态滤波器对复杂变化的图像具有良好的滤波性能和稳健的适应能力。
This paper presents morphological neural networks of opening and closing operation for pratical use, and the algorithm to design optimal parameters of a morphological filter, Experimental results show that this method is good in practice and easy to extend. It has better filtering properties than that of the conventional morphological ones. The task of creating a morphological filter can be divided into two basic problems, selecting a morphological operation and Structuring Element (SE). The set of morphological operations is predefined so the filter's properties depend merely on the selection of an SE. Structuring elements are formed by means of an adaptive algorithm that adjust the shape of the SE to match characteristics of the image targets. Morphological filters formed using this method are capable of responding complicated patterns in images.
出处
《电子与信息学报》
EI
CSCD
北大核心
2001年第11期1220-1224,共5页
Journal of Electronics & Information Technology
