摘要
基于迭代自组织数据聚类阈值的脉冲耦合神经网络的图像分割算法改进了传统脉冲耦合神经网络在图像分割中由于不恰当的参数选择而导致图像欠分割和过分割的问题。基于迭代自组织数据聚类阈值的脉冲耦合神经网络图像分割算法无需确定参数和循环次数,也不需要用特定原则确定循环结束的条件,只需利用图像中的每个像素点的灰度值进行聚类,然后利用改进的迭代自组织数据算法确定图像的初始聚类数目以及聚类中心,并以此作为脉冲耦合神经网络的最佳阈值,一次点火过程自动完成分割。实验结果表明,这种算法具有较好的分割结果和分割速度,提高了分割的准确性。
A novel image segmentation method based on ISODC - PCNN, is put forward to solve the problem of PCNN with improper parameter results in short of segmentation or over - segmentation. The proposed algorithm does not need to determine the model parameter, iteration time, or the iteration stop condition. ISODC - PCNN makes use of the grey level of the image to cluster, uses the improved ISODATA to determine the initial number and the center of clustering, which can be used as the optimum threshold value of PCNN. ISODC - PCNN can segment an image with one time of iteration. The result of the experiment shows that the proposed method has good performance and it is faster and more accurate than some other PCNN based on segmentation algorithms.
出处
《河池学院学报》
2014年第2期71-76,共6页
Journal of Hechi University
关键词
脉冲耦合神经网络
迭代自组织数据聚类
图像分割
Pulse Coupled Neural Network
Iterative Self- Organizing Data Clustering
image segmentation
