2024年6月14日发(作者：)

复杂物体轮廓提取

徐晓刚 于金辉 马利庄

(浙江大学CAD&CG国家重点实验室,杭州 310027)

摘要 图象分割是图象处理中的一项重要工作，目前手工与自动相结合的分割方法在实际工

作中得到了广泛应用。本文根据图象经Maar变换后的特征，采用新的判断准则，提出了一种复

杂物体边缘定位算法，可以对具有尖角特征的物体轮廓进行快速准确地提取，同时利用矢量化方

法消除毛刺，使跟踪获得的边界更符合物体的实际轮廓特征。对多种图象的实验表明本文方法十

分有效。

关键词 复杂轮廓提取，Snake 算法，基于动态规划图搜索算法

中图法分类号：TP751.1

Extraction of Complex Object Contour

XU Xiao-gang YU Jin-hui MA Li-zhuang

(State Key Lab of CAD&CG Zhejiang University Hangzhou 310027)

Abstract Image segment plays an important role in the field of image processing,

and currently the hybrid approach combining the manual and automatic methods is

widely used in segment practice. In this paper we present an algorithm capable of

本课题得到国家自然科学基金

(69973043)

与

(60073024)

资助。

locating the target object contour of sharp tips accurately in the interactive rate.

Considering that the edge are usually on the zero-crossing points after Marr

transformation for most images, existing techniques tend to give undesirable results

because the energy path containing more points is given less priority. In our method we

specify a pointer to point the current point on a path of interest, when the energy of

current path is less than the energy for the previous point, we check

n

latest points in the

current path instead of checking only one point as existing techniques do, and, if more

than

m

point (m≤n) is zero-crossing, the pointer of the point is updated, otherwise, the

pointer remains unchanged. Using this criterion we can insert new seeds automatically

near the tips of the target object and the burr is eliminated by a vectorization approach.

The final contour traced out fits the feature of the target object well and the effectiveness

of our method is demonstrated by examples shown in the paper.

Keywords contour detection, snake, graph searching formulation of dynamic

programming.

0引言

图象分割是一项广泛应用的图象处理技术。由于图象的多义性和复杂性，许多分割的工作无

法依靠计算机自动完成，而手工分割又存在工作量大，定位不准确的难题，因此，人们提出了一

些人工交互和计算机自动定位相结合的方法，利用各自的优势，实现目标轮廓的快速定位。纵观

这些方法，它们大致可以归结为两类：

一类为Snake 算法或Active Contour Models[1][2]；这类算法需要给出初始的轮廓，然后

进行迭代，使轮廓沿能量降低的方向靠近，最后得到一个优化的边界。能量函数包括内外力两方