摘要：岩心颗粒的彩色图像包含的信息具有复杂性和多样性,除了人眼视觉系统容易感知的颜色与空间形状特征之外,还隐含着更深层次的纹理特征信息。提出一种多维特征核模糊C均值(Kernel Fuzzy C-Means,KFCM)聚类分割算法:首先使用Gabor滤波器组在频域的不同尺度和方向上对岩心颗粒彩色图像进行卷积滤波处理,并将Gabor滤波结果作为频谱的局部纹理特征;然后将纹理特征、颜色特征以及图像像素点空间位置信息合并到核模糊C均值聚类算法中,从而实现岩心颗粒彩色图像的分割。结果表明:与其他算法的分割结果相比,多维特征KFCM聚类分割算法能更准确地识别不同类型的岩心颗粒的彩色图像,获得了良好的分割结果。

Abstract: Color images of core particles contain complex and diverse information．In addition to the color and spatial shape features easily perceived by human visual system，deeper texture feature information is also implied． In this paper， a clustering segmentation algorithm of Kernel Fuzzy C-Means ( KFCM) was proposed based on multi-dimensional features．Firstly，the color images of core particles were convoluted by Gabor filter banks at different scales and directions in the frequency domain，and the results of Gabor filtering were used as local texture features of the spectrum．Then，texture features，color features and spatial location information of image pixels were merged into the Kernel Fuzzy C-Mean clustering algorithm to realize color image segmentation of core particles．The results show that compared with the segmentation results of other algorithms，the multi-dimensional feature KFCM clustering segmentation algorithm can more accurately identify the color images of different types of core particles and obtain good segmentation results．

Keywords: core particle; color image segmentation; Gabor texture feature; Kernel Fuzzy C-Means clustering algorithm; multi-dimensional features