结肠镜图像自动分割技术研究  

Study of automatic segmentation technique for colonoscopic images

作　　者：朱丹琦[1] 夏顺仁[1] 楼晓敏[2] 周乐平[2] 张国雄[2] 

机构地区：[1]浙江大学生物医学工程教育部重点实验室,浙江杭州310027 [2]杭州市中医院,浙江杭州310006

出　　处：《浙江大学学报（工学版）》2004年第7期821-825,共5页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(60272029);浙江省自然科学基金资助项目(M603227).

摘　　要：提出了一种融合颜色、亮度、空间距离和纹理等特征的彩色结肠镜图像分割新算法.纹理特征采用一组Gabor滤波器对原始图像滤波后计算得到的分形维特征.利用基于元素间相似性的随机聚类方法对特征空间进行聚类.通过对图中的切割进行采样,自动获得最佳的类别数目,算法复杂度较低,其随机特性使得它具有抗噪声的鲁棒性.对多幅结肠镜图像进行分割实验,结果证实了该算法的有效性.An Algorithm for segmenting color colonoscopic images by fusing color, brightness, spatial distance and texture information was presented. The fractal dimensions, which served as the measurement for the texture features, were obtained from the filtered images of original images by a set of Gabor filters. A stochastic clustering method based on pairwise similarity of elements was utilized to cluster feature space. Based on the sampling of cuts in graphs, the optimal number of clusters was obtained automatically and with low complexity. Furthermore, the method's stochastic nature makes it robust against noise. The segmentation of several colonoscopic images proves the effectiveness of the algorithm.

关 键 词：图像分割 特征融合 分形维 元素间相似性 随机聚类 

分 类 号：TN911.73[电子电信—通信与信息系统] TP317.4[电子电信—信息与通信工程]