摘要
针对涎腺超声图像斑点噪声强、对比度低和边界弱的特点,提出了一种结合形态学检测的自动随机游走分割方法。该方法首先利用形态学操作获得目标的初始轮廓,然后提取目标区域和背景区域骨架结构的有效标记点作为随机游走算法的种子点,最后利用种子点对预滤波后的肿瘤图像实现随机游走分割。实验选取大量临床采集的涎腺肿瘤超声图像进行测试,结果表明该方法计算复杂度低,解决了传统随机游走模型初始种子点的人工干预问题,有效实现了涎腺肿瘤的自动分割。
In view of the difficult segmentation of ultrasonic tumor image with strong speckle noise,low contrast and weak boundaries,an automatic segmentation algorithm is proposed,combined random walk with morphology detection.Firstly,the rough contour of the target is obtained through morphological operation,then the skeleton structures of foreground regions and background regions are extracted,and the gauge points in the skeleton structures are labeled as the seed points for random walk.Finally,segmentation results are obtained by random walk with the labeled seeds on the filtered tumor images.The proposed method has been tested with large number of clinical salivary gland ultrasound images and the test results demonstrate that the proposed method has low computational complexity and overcomes the limitation of locating initial seed points manually in traditional random walk,thus realizing the automatic segmentation of salivary gland tumor effectively.
出处
《数据采集与处理》
CSCD
北大核心
2013年第5期553-558,共6页
Journal of Data Acquisition and Processing
基金
国家自然科学基金(61201363
61271305)资助项目
高等学校博士学科点专项科研基金(20110009110001)资助项目
中央高校基本科研业务费专项资金(2012JBM012
2010JBZ010)资助项目
北京交通大学人才基金(2012RC036)资助项目
关键词
超声图像分割
随机游走
涎腺肿瘤
斑点噪声
形态学
ultrasonic image segmentation
random walk
salivary gland tumor
speckle noise
morphology
