论文部分内容阅读
提出了一种基于视觉感知特性及改进的模糊Kohonen聚类网络的图象纹理分割方法，它由2D 最佳正交极可分方向滤波器特征提取，特征图象的四叉树平滑，改进的模糊Kohonen聚类网 络(IFKCN)特征聚类及边缘确定四部分组成。最后给出了实验结果。“,”To our best knowledge, there does not exist any segmentation method based both on visual perception and FKCN(fuzzy Kohonen clustering network). Our method is new in three respects: (1) it is based on a combination, not just a mixture, of vi sual perception and IFKCN (improved FKCN); (2) FKCN is improved; (3) a learning algorithm raises the speed of segmentation.rn We use the 2D polar separable filters proposed by the second author et al［ 1］ to do orientational filtering. In section 3, we discuss how to improve FKCN by using eqs.(4) through (9). Simulation experiments (results shown in Fig.3) show the effectiveness of our method.