基于区域生长法的医学图像分割研究  

Study on medical image segmentation based on regional growth method

作　　者：刘喜英[1] 吴淑泉[2] 徐向民[2] 

机构地区：[1]广东工业大学信息工程学院,广东广州510643 [2]华南理工大学信息学院,广东广州510640

出　　处：《电视技术》2003年第9期24-25,31,共3页Video Engineering

基　　金：广州市科技攻关项目(编号:200223-D3041)资助

摘　　要：针对基于区域的分水岭分割算法中通常存在的严重过分割现象,通过两点措施来改进:其一,以梯度图像的浮点活动图像取代梯度图像进行分水岭变换,使边缘定位更准确;其二,以基于面积和对比度控制的合并小区域准则,有效抑制过分割现象,同时满足感兴趣区的分割要求。实验表明该算法简单有效,能够得到符合人类视觉系统特性的分割结果,边缘定位精确度较高。To deal with the extreme over-segmentation problem of region-based watershed methods,two approaches have been proposed.Firstly,we perform the watershed transform on the floating-point moving image in stead of the morphological grad image,which can locate the contour more accurately.Secondly,we propose the region merging rules based on the area and contrast,which can restrain the over-segmentation problem effectively and reserve the interesting regions.The experimen-tal results demonstrate the simplicity and affectivity of the approach,and the results are accordant to our understanding on the semantic concepts and the perceptual mechanism of human eyes.

关 键 词：区域生长法 分水岭算法 过分割现象 形态梯度图像 浮点活动图像 合并小区域 图像分割 

分 类 号：TN911.73[电子电信—通信与信息系统] R445[电子电信—信息与通信工程]