一种基于Hellinger距离的活动轮廓图像分割方法

摘要

As the basis of lots of computer vision applications,image segmentation has been the research focus of the scholars in this sector over past decade.In this paper we present a CV active contour model-based image segmentation method,it uses Hellinger distance to measure the differences of a region with each pixel within it to replace the variance measure in CV active contour.Combining with the idea of geodesic active contour (GAC),we improve the measures of region boundary length in objective function of CV model.Then the objective function is mapped onto network flow graph and to be optimised by GraphCuts.A large number of experiments show that the proposed method has better segmentation results and has been greatly improved in contrast with CV active contour method.%图像分割作为许多计算机视觉应用的基础，近十年来一直是计算机视觉学者们研究的焦点。提出一种基于CV活动轮廓模型的图像分割方法，采用Hellinger距离度量区域内的每个像素点与区域的差异度，取代CV活动轮廓中的方差度量，并结合测地活动轮廓（GAC）的思想，改进了CV模型目标函数中区域边界长度的度量。然后将目标函数映射到网络流图中，采用图切割方法对目标函数进行优化。实验证明所提出的方法具有较好的分割效果，相对于CV活动轮廓法有了较大的提升。