基于快速高斯核函数模糊聚类算法的图像分割

Image Segmentation Based on Fast Gauss Kernel Function Fuzzy Clustering Algorithm

对模糊聚类算法通过引入高斯核函数,平滑图像像素灰度值,从而增强图像分割的抗干扰能力和鲁棒性,并结合阈值模糊聚类算法,提高了图像分割的速度。首先利用阈值模糊聚类法划分初始输入空间,得到模糊规则数及初始聚类中心;然后用高斯核函数平滑图像的像素灰度值;最后用标准模糊聚类算法求解并优化模糊隶属度和聚类中心。将本算法应用于添加噪声的嫦娥一号采集的月球地面灰度图像和Lena灰度图像进行图像分割,仿真结果验证了本方法的鲁棒性、有效性和实用性。

Gauss kernel function was introduced to fuzzy clustering algorithm to smooth the image pixel gray value,thereby the anti-jamming capability and robustness of the image segmentation was enhanced.The speed of image segmentation was improved by combining with the threshold value fuzzy clustering algorithm.First of all,the algorithm started from an initial fuzzy partition of input space by threshold value fuzzy clustering algorithm to get the number of fuzzy rules and the initial clustering center;then the gray level of image was smoothed by Gauss kernel function;at last the fuzzy membership and the clustering center were optimized with the original fuzzy clustering algorithm.The proposed algorithm was applied to the image segmentation of Chang E-1 acquisition gray image of the moon surface and Lena gray image which joined with noise.Simulation results demonstrate the robustness,effectiveness and practicality of the proposed fast Gauss kernel function fuzzy clustering algorithm.