CAJ | 학술논문

针对传统的模糊C均值聚类算法在进行图像分割时对孤立点、噪声点敏感性较强，聚类耗时随图像变大而快速增长等缺陷，基于临近元素空间距离的模糊C均值聚类算法即SFGFCM算法，采用核化的空间距离公式，计算出空间临近像素与考察像素的相似度Sij ，然后用邻近像素灰度加权和计算出邻近信息制约图像，并进一步在邻近信息制约图像的灰度级统计的基础上进行聚类。该算法考察了临近像素灰度和位置等信息，并且它们之间取得了很好的平衡；不仅表现出较强的鲁棒性且很好地保留了原图像边缘等细节信息，提高了聚类精度，同时大大缩短了大幅图像的聚类时间。通过在合成图像、医学图像及自然图像上的大量实验，与传统算法对比该算法聚类性能明显提高，在图像分割上体现出了较好的分割效果。
침대전통적모호C균치취류산법재진행도상분할시대고립점、조성점민감성교강，취류모시수도상변대이쾌속증장등결함，기우림근원소공간거리적모호C균치취류산법즉SFGFCM산법，채용핵화적공간거리공식，계산출공간림근상소여고찰상소적상사도Sij ，연후용린근상소회도가권화계산출린근신식제약도상，병진일보재린근신식제약도상적회도급통계적기출상진행취류。해산법고찰료림근상소회도화위치등신식，병차타문지간취득료흔호적평형；불부표현출교강적로봉성차흔호지보류료원도상변연등세절신식，제고료취류정도，동시대대축단료대폭도상적취류시간。통과재합성도상、의학도상급자연도상상적대량실험，여전통산법대비해산법취류성능명현제고，재도상분할상체현출료교호적분할효과。
To deal with the traditional image segmentation algorithms’sensitive to noises and outliers, and the segmenting time increasing with the image size, the fast and Fuzzy C-Means clustering algorithm based on the space distance of near-est neighbors(SFGFCM)uses a nuclear space distance formula to calculate the similarity measure Sij between the query pixel and its nearest neighbors, further more to calculate the local information constrains image with the adjacent pixels weighted sum, at last sum the local information image’s grey level to cluster. So that it well balances the query pixel’s gray levels and local information, provides robustness to noisy images and guaranteed image details presentation, im-proves the accuracy of clustering, at the same time reduces the clustering time remarkably for the large size images. Exper-iments performed on a large number of synthetic and real-world images show that SFGFCM is more effective and effi-cient in contrast to traditional algorithm on image segmentation.