二型模糊融合的尺度约束多分辨率FCM分割

针对传统多分辨率模糊聚类图像分割算法的不足,提出了将二型模糊应用于多分辨率模糊聚类图像分割的新方法.将最粗尺度图像的聚类中心作为下一较细分辨率图像的初始聚类中心,并采用较粗分辨率图像聚类的类内最大距离对细分辨率图像的模糊聚类目标函数进行约束.对较小的粗分辨率图像进行了模糊隶属度扩展,得到一组隶属度值,再采用二型模糊算法有效融合该隶属度集合,完成聚类分割.实验结果表明,该算法能有效实现目标区域分离,获得理想分割效果.

Multi-resolution FCM with scale constraints for segmentation using type-2 fuzzy fusion

In order to overcome the problem of traditional multi-resolution fuzzy clustering for image segmentation, a new method of multi-resolution fuzzy clustering for image segmentation based on type-2 fuzzy is proposed. The cluster center gotten by the low resolution image clustering, is the initial center for the higher resolution image. The object function of the higher resolution image is added by the maximum distances of intra-class in the low resolution image as a constraint of the low resolution image. The ...