基于边缘流和距离图Snake模型分割淋巴结超声图像  

Segmentation of Lymphatic Ultrasonic Images Based on The Edge Flow and Distance Map Snake Model

作　　者：张俊华[1] 汪源源[1] 施心陵[2] 董怡[3] 王怡[3] 

机构地区：[1]复旦大学电子工程系,上海200433 [2]云南大学电子工程系,昆明650091 [3]复旦大学附属华山医院,上海200040

出　　处：《中国生物医学工程学报》2006年第5期532-537,共6页Chinese Journal of Biomedical Engineering

基　　金：国家重点基础研究规划基金(2005CB724303);上海市曙光计划(2003-901)

摘　　要：提出一种基于边缘流的距离图Snake模型的图像分割方法,用于淋巴结超声图像的分割。首先由给定的4个标记点获得Snake模型的初始轮廓,然后综合图像灰度和纹理特征构造边缘流,由边缘流演化所得边缘图来构造距离图,通过定义基于距离图的势能函数,作为Snake模型的外部势能,来引导模型形变,实现对淋巴结超声图像的半自动分割。A modified distance map （DM） Snake model based on the edge flow was proposed to segment ultrasonic images of lymph nodes. Firstly, the contour of the Snake model was initialized according to the 4 marks provided by an expert. Secondly, an edge flow vector was constructed at each pixel based on the integration of image intensity and texture characteristics. The distance map was formed by the edge map, which was obtained through the evolution of the edge flow. Then, the potential energy was defined based on the modified DM as the extern energy of the Snake model. Finally, the curve propagation was directed by the modified DM snake model to semi-automatically segment the ultrasonic images of lymph nodes.

关 键 词：边缘流 距离图 SNAKE模型 医学超声图像 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] R319[自动化与计算机技术—计算机科学与技术]