一类具有连通性探测功能的CNN研究与应用

基于细胞神经网络(Cellular Neural Network,CNN)的图像处理的研究和应用已取得了很大进展。在图像处理中,有时需要确定图像是否具有某类特殊的连通性质。本文对CNN模板库列出的一个CNN模板进行了分析,确定出了它的局部规则,提出了一个满足该局部规则的CNN鲁棒性设计定理,它为设计相应的CNN模板参数提供了解析判据。通过数值模拟实例确认了理论结果在计算机图像处理中应用的有效性。

Design for CNN Templates with Performance of Global Connectivity Detection

The cellular neural/nonlinear network (CNN) is a powerful tool for image and video signal processing,robotic and biological visions. This paper discusses a general method for designing template of the global connectivitydetection (GCD) CNN, which provides parameter inequalities for determining parameter intervals for implementing thecorresponding functions. The GCD CNN has stronger ability and faster rate for determining global connectivity in binarypatterns than the GCD CNN proposed by Zarandy. An example for detecting the connectivity in complex patterns isgiven.