基于归一化割的血吸虫卵图像分割  

Segmentation of Schistosome Eggs Image Based on Normalized Cut

作　　者：叶超[1] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,江苏南京210016

出　　处：《计算机技术与发展》2015年第11期27-31,37,共6页Computer Technology and Development

基　　金：江苏省科技支撑计划(BE2013879);江苏省自然科学基金(BK20131365);江苏省第九批"六大人才高峰"高层次人才项目资助方案(DZXX-035)

摘　　要：显微镜下识别血吸虫卵是一件费时费力的工作,常常因为检测人员疲劳、注意力不集中等原因导致血吸虫病的误检和漏检。传统的血吸虫卵识别系统,图像分割方法只是通过单一的阈值分割的方式来处理血吸虫卵图像,往往误诊率、漏检率较高,在血吸虫病的诊断中作用很小甚至有副作用,因此对图像分割算法进行改进变得很有必要。文中在此背景下进行研究,针对血吸虫卵图像的分割提出了改进的归一化割算法。采用灰度权值矩阵描述像素之间的关系,从而避免了特征系统的大量运算,同时结合了先验知识,根据图像自身的特点自动计算最优分割子图数,使得分割结果更加精确。实验结果表明,该分割算法比经典的阈值分割算法分割更加精确,并且运行速度快。It is a laborious work to identify the schistosome eggs under microscope. Because of fatigue testing and inattention, testing personnel often causes false and leak detection. Traditional recognition system＇ s image segmentation method is simply using a single threshold, and the detection result is not satisfied normally. It is almost useless in the diagnosis of schistosome. So to improve the image segmen- tation algorithm becomes very necessary. In this paper, under the background of the study,in view of the schistosome eggs image segmen- tation, an improved normalized cut algorithm is proposed. Use gray weighting matrix to describe the relationship between the pixel, so as to avoid the operation of characteristics system, at the same time, combined with prior knowledge, according to the characteristics of the image itself automatically calculates the optimal segmentation figure number, which make segmentation result more accurate. The experimental result shows that the algorithm runs faster with higher accuracy than typical threshold segmentation algorithm.

关 键 词：医学图像 血吸虫卵 图像分割 归一化割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]