空间模糊C均值聚类的神经切片图像分割方法

Spatial information based fuzzy C-means for nerve sliced image segmentation

周围神经切片显微图像具有背景复杂、区域不连续和光照不均匀等特点,应用经典的图像分割算法难以取得有效的分割结果。通过结合初始隶属度概率函数和空间距离来设计空间函数而得到的SFCM聚类算法,并提出SFCM彩色图像分割方法。把图像从RGB颜色空间转换到HIS颜色空间。采用聚类有效性分析指标在直方图快速FCM算法中为HSI各分量确定分类数目和获取SFCM初始化参数。对HIS各分量分别进行SFCM聚类,合并各分量并转换回RGB彩色空间以显示结果。实验结果表明,与标准FCM聚类分割算法相比,新方法能更有效地分割区域不连续的神经切片显微图像。

Peripheral nerve sliced microscopic images have the characteristics of complex background, discontinu- ous regions and non-uniform illumination.It is difficult to apply classical image segmentation algorithms to obtain a valid segmentation result.By combining the probability of the initial membership functions and space distance to de- sign the space function of SFCM clustering algorithm, it proposes SFCM color image segmentation method in this paper.The image color space is converted from RGB to HIS color space.Cluster validity function is used to define the number of clusters of each component and initialize the algorithm with the algorithm of FCM based on image histogram.SFCM is applied separately on each component of HSI model, and each component is combined and dis- played in RGB model.The experimental results show that compared with the standard FCM clustering segmentation algorithm, the new method is more effective for segmenting discontinuous nerve sliced microscopic images.