基于改进模糊聚类算法的CT图像病变区域分割  

CT image′s lesion area segmentation based on improved fuzzy clustering algorithm

作　　者：苏博[1] 朱敏[2] 

机构地区：[1]贵州理工学院信息工程学院,贵州贵阳550003 [2]贵州大学农业生物工程研究院,贵州贵阳550003

出　　处：《现代电子技术》2016年第24期100-103,108,共5页Modern Electronics Technique

基　　金：贵州省科技厅自然科技基金:贵阳交通智能控制与诱导技术研究(qian J word[2012]2012)

摘　　要：针对CT图像病变区域存在的欠分割和过分割问题,提出基于融合空域滤波器的改进模糊聚类算法。对CT图像进行预处理,将CT图像变换成通用图像,使输入图像具有一致直方图灰度分布,用区域生长方法标记和过滤非目标病变区域,获取脑部CT图像目标病变区域。通过改进的空域滤波器对噪声图像进行滤波去噪,对模糊聚类算法的目标函数和迭代公式融入空域滤波数据项确定隶属度矩阵,完成CT图像病变区域的分割。实验结果表明,所提方法在人脑CT图像血块区域进行分割过程中,具有较高的分割效率和精度,对噪声具有较高的鲁棒性。To solve the under-segmentation and over-segmentation problems existing in the lesion area segmentation of CT images, an improved fuzzy clustering algorithm based on spatial filter fusion is put forward. The preprocessing of CT image can transform the CT image into a general image by means of the algorithm to make the input image have the consistent histogram gray-scale distribution. The region growing method is used to mark and filter the non-target lesion area to obtain the target lesion area of the brain CT images. The improved spatial filter is used to conduct the filtering and denoising of the noise image. The ob- jective function and iterative formula of the fuzzy clustering algorithm are added into the spatial filtering data items to determine the membership matrix, so as to segment the lesion area of CT images. The experimental results indicate that the algorithm has high segmentation efficiency and precision in the segmentation process of blood clot region of the human brain CT images, and good robustness to resist the noise.

关 键 词：改进模糊聚类算法 CT图像 病变区域分割 隶属度矩阵 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置] TP391[自动化与计算机技术—控制科学与工程]