支持向量学习并行区域增长结合活动轮廓模型的图像分割算法

为克服经典区域增长算法门限设置困难和图像分割精度不高的问题,提出了基于支持向量机学习的区域增长与活动轮廓模型结合的高精度图像分割算法。首先交互式选择属于目标区域的子块和背景区域的子块形成支持向量机的训练样本;并利用这些已知的训练样本训练支持向量分类器。在目标与背景的并行竞争增长过程中,利用训练好的支持向量分类器(SVC)进行分类判决,得到目标对象的初始轮廓。为提高分割对象的精度,采用活动轮廓模型获得准确的边缘。仿真实验获得了较好的分割效果,表明该提出的算法是合理可行的。

Accurate image segmentation based on support vector region growing and active contour model

To overcome the difficult of threshold initialization and precise object segmentation existing in conventional region growing algorithm,a accurate image segmentation method combined support vector learning parallel region growing and active contour model was proposed.The blocks belong to the object region or belong to the background region were interactive selected.The support vector classifier was trained by the training data collected in the first step.In region growing processing step,the support vector classifier(SVC) with maximum soft margin was used for establishing region growing rules,so as to obtain the initial object contour.In order to achieve accurate segmentation results,the active contour model was exploited to further segment to got accurate boundary. Experimental results show that this algorithm is feasible and it performs better than conventional region growing algorithm.