基于单水平集的多目标轮廓提取

多目标轮廓提取是图像分割的重要研究内容.文章在Chan和Vese的无边界主动轮廓模型(简称CV模型)的基础上,提出基于单水平集的多目标轮廓提取算法.CV模型只能实现单目标的轮廓提取,主要原因是不能使水平集函数驱动的轮廓线在某些目标区域正确分裂,没有有效利用轮廓线的拓扑分裂信息.通过修正CV模型,引入标记模板,用于追踪零水平集的分裂情况,对不同的准目标区域区别处理;引入图像区域均值模板,用于估计可能淹没在背景区域中的目标区域,促使水平集函数在上述目标区域充分变形,使对应零水平集充分分裂,实现多目标轮廓提取.并且文章提供了一系列不同条件下的实验结果,并与其它类似的研究成果进行比较,结果表明,该文的工作是有意义的.

Multi-Objects Extracted Based on Single Level Set

Multiple-objects contour extraction is an important area in image segmentation.Based on the active contours model without edges,known as CV model,which is proposed by Chan and Vese,the authors propose a multiple-objects contour extraction algorithm based on single level set function.CV models just shed the light on single object contour extraction.The major cause is it makes the level-set-function driven contours to split in some targeted areas,which do not make use of topology split information of the contours effectively.By modifying the CV model,this paper introduces the mark template to track the split information of the zero-level-set and carries out different scheme at different targeted areas.Furthermore,this paper introduces image area average template to evaluate the possible targeted area submerged in the background,which deform the level set function efficiently on the above targeted area to realize the zero-level-set split and multiple-objects extraction.The authors have offered a series of experimental results under different conditions and have compared with other research results in this work.