微分进化自适应模糊C均值分割算法

发布时间：2018-05-10 15:26

本文选题：模糊C均值 + 图像分割　；参考：《计算机工程与应用》2017年23期

【摘要】：模糊C均值(FCM)聚类算法分割图像时,对图像的背景噪声和聚类算法的初始值比较敏感,为了克服这个问题,进而提出了微分进化模糊C均值分割算法。为了避免陷入局部极值,首先使用FCM聚类初始化,接着用改进的FCM进行模糊聚类;然后进行初始化种群操作,设置微分进化DE算法的参数,计算种群中每个个体的适应值,最后对满足条件的适应值进行变异、交叉、选择操作。利用DE算法的全局搜索优化能力,有效抑制了局部极值的产生和图像的背景噪声、纹理细节对图像分割效果的影响。还克服了对初值选择敏感的问题,保证图像分割边界的完整性,是一个比较高效的方法,有效地提升了分割效果。DE算法本身具有简单,快速,鲁棒性好等优点,利用这些优点可以有效地克服FCM算法的缺点。
[Abstract]:The fuzzy C-means (FCM) clustering algorithm is sensitive to the background noise of the image and the initial value of the clustering algorithm. In order to overcome this problem, a differential evolutionary fuzzy C-means segmentation algorithm is proposed. In order to avoid falling into local extremum, FCM clustering is first used to initialize, then fuzzy clustering is performed with improved FCM, then the population operation is initialized, the parameters of differential evolution DE algorithm are set, and the fitness value of each individual in the population is calculated. Finally, the adaptive value of the condition is changed, crossed, and selected operation. The global search optimization ability of DE algorithm is used to effectively suppress the local extremum and the background noise of the image, and the effect of texture details on the image segmentation effect. It also overcomes the problem of sensitivity to initial value selection and ensures the integrity of image segmentation boundary. It is a more efficient method and effectively improves the segmentation effect of .DE algorithm, which has the advantages of simple, fast and good robustness. These advantages can effectively overcome the shortcomings of FCM algorithm.
【作者单位】：西安理工大学自动化与信息工程学院;
【基金】：国家自然科学基金重大项目课题四(No.41390454)
【分类号】：TP391.41

【相似文献】