基于改进型聚类神经网络的图像分割

Image segmentation based on improved clustering neural network

将聚类网络用于非监督的图像分割,提出了竞争层神经元的动态调整机制和返回式的非重复训练学习方案,实现了聚类数的自适应增加,解决了随机生成权值矩阵产生的死点问题,提高了算法的收敛性能。实验结果表明,改进的聚类网络的图像分割结果优于C-均值聚类算法和通常的聚类网络。

This paper applies the clustering neural network to unsupervised image segmentation.The automatic mechanism of nerve cell and recurrent non-repeating training scheme are proposed to implement the adaptive adding of the clustering number, solve the dead center of stochastic weighted matrix,and improve the convergence.Experiments show that our improved clustering neural network is better than C-mean algorithm and the accustomed clustering network to segment images.