基于改进的区域生长与GVF snake的超声图像分割  

Segmentation of Ultrasound Images Based on Improved Region Grow and GVF Snake

作　　者：刘丹[1] 杨艳[1] 孙尽尧[1] 

机构地区：[1]武汉大学物理科学与技术学院,武汉430072

出　　处：《科技通报》2016年第4期165-169,共5页Bulletin of Science and Technology

基　　金：国家重点基础研究发展计划(973)项目(2011CB707900)

摘　　要：针对传统的区域生长法应用于超声图像时,在选取相邻的生长阈值的情况下会分别出现欠分割与过分割而导致分割失败的现象,提出一种改进方案,在生长之前先将每个待生长像素进行压缩之后再与种子点进行比较,并通过改进的方案与各向异性滤波技术的结合获得超声图像目标区域并提取其轮廓,然后将此轮廓作为GVF Snake模型的初始轮廓收敛获得最终的分割结果。通过对临床采集的超声图像进行的实验结果表明,改进后的区域生长法能成功地应用于超声图像,并能很好地作为GVFSnake的初始模型从而获得理想的分割结果。An improved region grow for the growth rule is introduced effectively to overcome the under-segmentation and over-segmentation of which when there are neighbouring threshold values in the traditional region grow which is applied to ultrasound images.Then as the combination of the improved approach and anisotropic diffusion, an initial coutour can be gained for GVF snake to converge to a final segmentation.The results of the experiment on the clinical sample show that the improved approach can apply successfully to ultrasound images,and it’s also can be a good start for the GVF snake to draw a final desired segmentation.

关 键 词：区域生长法 活动轮廓模型 超声图像分割 GVF 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]