模糊Kohonen聚类网络用于骨髓细胞图像分割

Fuzzy Kohonen clustering network for medullar cell image segmentation

提出了一种基于图像的边界特点来确定模糊Kohonen聚类网络输出类别的方法 .该方法首先构建了图像的一个边界模型 ,然后利用Zernike矩求出边界模型的参数 ,最后根据参数的特点 ,自动确定模糊Kohonen聚类网络输出数目 .试验结果表明该方法构建的模糊Kohonen聚类网络能够实现图像的自动模糊分割 。

A new method based on image's boundary characteristic is presentedto determine the numbers of fuzzy kononen clustering Network's output nodes. In this method, an image's boundary model is presented, and the model's parameters are acquired by using Zernike moments. The FKCN's output numbers are then confirmed automatically based on the traits of the parameters. The result shows that the FKCN made by the proposed method can segment images automatically and therefore it is feasible.