CAJ | 학술논문

针对高噪声、低对比度的医学图像难以快速准确分割的问题,结合基于像素的传统方法和基于水平集的活动轮廓模型,提出了一种混合的医学图像分割新技术.首先依据待分割对象的先验知识交互选取感兴趣区域.然后由传统的方法和基于水平集的C-V模型结合实现感兴趣区域图像的预分割.预分割的结果直接作为窄带变分水平集模型的初始轮廓,演化曲线在很短的时间内准确收敛到待分割物体的边缘.
침대고조성、저대비도적의학도상난이쾌속준학분할적문제,결합기우상소적전통방법화기우수평집적활동륜곽모형,제출료일충혼합적의학도상분할신기술.수선의거대분할대상적선험지식교호선취감흥취구역.연후유전통적방법화기우수평집적C-V모형결합실현감흥취구역도상적예분할.예분할적결과직접작위착대변분수평집모형적초시륜곽,연화곡선재흔단적시간내준학수렴도대분할물체적변연.
In view of the difficulty in accurately and fast segmentation of high-noise and low-contrast medical image, a hybrid segmentation method, which combines traditional pixel-based method with level-set-based active contour model, is proposed. First, region of interest is interactively selected according to the pre-knowledge of object size. Then, pre-segmentation is done undergone the corporate result of traditional method and region-based C-V model. The pre-segmentation result is directly used as the initial contour of the narrow band based variational level set method. The evolving curve converges to object boundaries in a relatively short time.