一种新的用于医学图像分割的几何活动轮廊模型(英文)  

A NEW GEOMETRIC ACTIVE CONTOUR FOR MEDICAL IMAGE SEGMENTATION

作　　者：岑峰[1] 戚飞虎[1] 

机构地区：[1]上海交通大学计算机科学与工程系,上海200030

出　　处：《红外与毫米波学报》2003年第6期441-446,共6页Journal of Infrared and Millimeter Waves

基　　金：国家自然科学基金 (批准号 60 0 72 0 2 9)资助项目~~

摘　　要：提出一种新的几何活动轮廊模型对医学图像进行分割 .首先 ,我们对几何活动轮廊模型中吸引力场进行正则化 ,扩大目标轮廊边缘对的轮廊曲线的吸引力范围 ,增加轮廊曲线搜寻凹轮廊的能力 .然后 ,采用多尺度模型增加对边缘提取的精确度 .将正则化方法与多尺度方法相结合 ,能够很好的抑制医学图像中的噪声和虚假边缘的干扰 .这一方法能够在不采用任何附加拓扑控制的情况下自动控制轮廊曲线的拓扑结构变化 ,同时提取多个解剖结构 .对来自不同成像技术的医学图像的分割 ,结果表明该方法是一种有效的医学图像分割方法 .Generally, the segmentation of a medical image is difficult, because the medical image is often corrupted by norrupted by noise, and the anatomical shape in the medical image is complicated. In this paper presents a new geometric active contour scheme for medical image segmentation. First, we regularize the attraction force field in the geometric active contour model to extend the capture range of the object boundaries, and improve the ability of convergence to the concavities. Then, using a multi-scale scheme improve the boundary detection accuracy. In addition, combining the regularization and the multi-scale method, the proposed scheme can effectively suppress and eliminate the noise and the spurious edges in the medical images. Furthermore, the topology of the deforming curve can naturally change without and special topolygy handing procedures added to the scheme. This permits synchronously extracting several anatomical structures. The experiments on some medical images obtained from different medical imaging methods demonstrate that the proposed approach is competent for medical image segmentation.

关 键 词：医学图像分割 几何活动轮廊 Level-Set理论 边缘提取 噪声 虚假边缘 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]