基于Snake模型的快速边缘检测

摘要

There are two difficulties with Snake, one is sensitive to initialization and the other is hard to deep into concave boundary regions.Series of improved Snake are able to deal with above two difficulties.This paper puts forwards a rapid edge detection method which combines the advantages of Distance Snake and GVF Snake.First, approximate the object contour preliminarily by Distance Snake and design Criterion to determine the contour approximation.Then, using the GVF Snake to drive the contour into the concave regions while the contour approximation has been surpassed the given value.The results show that this new model not only has a large capture range and fast convergence, also could deep into concave regions and obtain the intact object contour.%传统Snake模型存在两个难点,一是初始轮廓敏感,二是难深入凹陷区域.针对这两问题,存在一系列改进模型.本文结合距离模型和GVF模型的优点,提出一种快速边缘检测方法.先采用距离模型快速逼近目标边缘轮廓,设计判别条件,判断逼近程度;当判断已经收敛到目标轮廓处时,利用GVF模型继续收敛,深入目标轮廓的凹陷区域.实验结果表明改进模型具有捕获区域大,收敛速度快以及能深入凹陷区域,检测出完整边缘轮廓的特点.