一种基于空间矩的PCNN图像分割方法*

脉冲耦合神经网络在图像分割中需要选取适当的网络参数，就内部连接矩阵而言，不同场合一般采用相同的设置，为了进一步改善图像分割质量，充分利用邻域的边缘及灰度分布信息，提出了一种基于空间矩确定内部连接矩阵的方法。仿真实验证明，该方法使得分割结果细节更丰富、分割区域更完整、边缘连续性更好。

Approach to image segmentation with spatial moments based on PCNN

The performance of pulse coupled neural network (PCNN) based algorithm for image segmentation depended on the suitable parameters setting for PCNN. For linking input matrix, it was mainly set fixed. In order to improve the quality of PCNN segmentation, proposed a novel method, based on spatial moments extracted from the grey values of some local area of an image, for the settings of linking input matrix, which took good use of the information of edge and grey values. Experiments show that with the proposed method, more details are illustrated in the segment results, and the edges and objective areas are more continuous and integral.