结合差分曲率的空间模糊C均值图像分割算法

Spatial Fuzzy C-Means Clustering Incorporating with Difference Curvature for Image Segmentation

针对基于空间信息的模糊C均值图像分割算法(sFCM)在对含噪图像分割时,图像的噪声和边缘细节不能同时得到较为正确分割的问题,本文提出了一种结合差分曲率的改进sFCM算法.差分曲率(difference curvature)可以有效地区分图像边缘和平坦区.将差分曲率引入到sFCM算法的空间函数中,算法的函数相关性参数在每个像素点处自适应取值,使改进算法在抗噪性能提高的同时,对图像细节有着更好的分割效果.实验结果表明:在对含噪图像进行分割时,本文提出的改进算法相比于sFCM及其衍生算法具有更好的模糊划分效果,并有效地提升了sFCM算法的抗噪性和对边缘细节的保护能力.

For the problem of noise and details can not be segmented correctly at the same time when applying spatial information fuzzy c-means clustering algorithm（sFCM）to noise image segmentation,an improved algorithm of sFCM incorporating with difference curvature was proposed in this paper.Difference curvature can distinguish edges from ramp regions effectively.By applying difference curvature to the spatial function of sFCM,the function correlation parameters are determined adaptively in a single pixel so that the adapted method is more robust to noise,and results in better segmentation performance for details in the meantime.The experimental results indicate that the improved method achieve competitive results in fuzzy clustering,compared to sFCM and its variants.The noise tolerance and detail-preserving property are promoted effectively.