完全基于边缘信息的目标靶快速分割算法

摘要

为了快速准确地分割提取目标靶,实现实时处理要求,提出了完全基于边缘信息的目标靶快速分割算法.首先采用基于邻域权值的非极大值抑制法细化Canny算子检测的边缘,并引入断点检测思想,通过构建追踪区域实现断点边缘追踪及连接,而后通过设置四向扫描长度总阈值,对边缘像点实行四向同步扫描填充,最终实现目标区域的准确分割.实验结果表明:相比其他分割算法,本文算法在分割正确率上具有较大优势,且采用较强的并行运行策略,提高了运行效率,达到了目标靶分割提取的实时性要求.%In order to segment and extract the target fast and accurately,a new fast segmentation algorithm fully based on edge information is proposed.Firstly the edge detected by Canny operator is thinned by the non-maxima suppression method based on the neighborhood weight.And then the breakpoint detection is introduced,the breakpoint edges are traced and connected by building tracing region.Finally the edge pixels are scanned and filled synchronously in four directions with the global threshold of the four-directional scanning length,and the accurate segmentation of the target is achieved.Experiment results show that compared with other algorithms,the proposed algorithm is of great advantage in segmentation accuracy,and is of high operating efficiency by using strong parallel strategy,and reaches the realtime requirement of the target segmentation.