改进的基于灰度级的模糊C均值图像分割算法

Improved fuzzy C-means algorithm based on gray-level for image segmentation

基于灰度级的模糊C均值算法是一种快速的图像分割算法。因为无损检测图像灰度分布不均衡,该算法不能有效分割图像中的目标与背景,故提出一种改进的基于灰度级的模糊C均值算法(IFCMG)。首先,利用灰度级像素数和隶属度构造类的总隶属度表达式并将其融入目标函数中以均衡化目标像素和灰度像素对目标函数的贡献。接着,推导基于新目标函数的隶属度和聚类中心。然后,考虑到类的密度也会影响聚类结果,设计类的紧密度表征形式并将其融入聚类进程。最后,采用无损检测图像进行分割实验。对于每幅图像,本文算法具有较高的F_value指标值。利用综合评价公式对所有F_value值进行评价,本文算法综合评价值比对比算法分别高出26.13%,16.46%,13.75%,25.10%。本文算法能够有效分割具有灰度分布不均衡特征的无损检测图像,扩展了基于灰度级的模糊C均值聚类算法的应用范围。

Fuzzy c-means algorithm based on gray-level is a fast image segmentation algorithm,which cannot effectively segment the object pixels and background pixels of the non-destructive testing(NDT)image with the characteristics of unbalanced gray distribution.Then an improved fuzzy C-means algorithm based on gray-level(IFCMG)is proposed.Firstly,the expression of total membership degree of each cluster is constructed by using pixel numbers and membership degrees of gray-level,and it is integrated into the objective function,which can equalize the contribution of the object pixels and background pixels to the objective function.Secondly,the new membership degree and cluster center are strictly deduced.And then,considering that the density of clusters also affects the clustering results,we design the formula of compactness and integrate it into the clustering process.Finally,the NDT images are used for segmentation experiment.For each image,IFCMG has higher index values of F_value when the images are disturbed by different noise levels.We comprehensively evaluate the values of F_value obtained above,and find that the comprehensively evaluation value of the proposed algorithm is 26.13%,16.46%,13.75%and 25.10%higher than those of the comparison algorithms,respectively.The proposed algorithm can effectively segment NDT images with unbalanced gray distribution,which expands the application scope of fuzzy C-means algorithm based on gray-level.