Edge contours using multiple scales

Donna J Williams, Mubarak Shah
{"title":"Edge contours using multiple scales","authors":"Donna J Williams,&nbsp;Mubarak Shah","doi":"10.1016/0734-189X(90)90003-E","DOIUrl":null,"url":null,"abstract":"<div><p>An algorithm for finding a single good path through the set of edge points detected by gradient of Gaussian operator is discussed. First, an algorithm for finding contours at one scale is presented, then an extension of that algorithm which uses multiple scales to produce improved detection of weak edges is presented. The set of possible edge points is placed on a priority queue with the edge point having largest magnitude at the top. The strongest edge point that is not already on a contour is retrieved from the queue. The point in the computed direction is examined first, then in those in the adjacent directions on either side of it. Each branch is followed to the end and a weight assigned at each point based on four factors: a measure of noisiness, a measure of curvature, contour length, and the gradient magnitude. The point with the largest average weight is chosen. After searching from the initial point in one direction, a similar search is conducted in the oppositedirection unless a closed contour has been formed. In the algorithm for multiple scales the search for a contour proceeds as for the single scale, using the largest scale, until a best partial contour at that scale has been found. Then the next finer scale is chosen and the neighborhood around the end points of the contour are examined to determine possible edge points in a direction similar to the end point of the contour. The original algorithm is then followed for each of the points satisfying the above condition, and the best is chosen as an extension to the original edge. Further, in order to determine the size neighborhood that should be searched when attempting to pick up an edge at a smaller scale, a theoretical analysis of the movement of idealized edges is performed. This analysis examines two adjacent step edges having the same parity (a staircase) and opposite parity (a pulse). It is determined that maximum movement for both cases is σ, where σ is the standard deviation of the Gaussian used. This maximum movement occurs for the staircase when the two nearby edges have the same step size and are at a distance of 2σ apart. However, for edges closer or farther away, maximum movement decreases rapidly. For a pulse, maximum movement occurs when the two edges have the same step size and are very close together. Again the movement decreases rapidly as the edges become farther apart. Movement also decreases in both cases when the relative strengths of the two edges are not equal.</p></div>","PeriodicalId":100319,"journal":{"name":"Computer Vision, Graphics, and Image Processing","volume":"51 3","pages":"Pages 256-274"},"PeriodicalIF":0.0000,"publicationDate":"1990-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1016/0734-189X(90)90003-E","citationCount":"69","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computer Vision, Graphics, and Image Processing","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/0734189X9090003E","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 69

Abstract

An algorithm for finding a single good path through the set of edge points detected by gradient of Gaussian operator is discussed. First, an algorithm for finding contours at one scale is presented, then an extension of that algorithm which uses multiple scales to produce improved detection of weak edges is presented. The set of possible edge points is placed on a priority queue with the edge point having largest magnitude at the top. The strongest edge point that is not already on a contour is retrieved from the queue. The point in the computed direction is examined first, then in those in the adjacent directions on either side of it. Each branch is followed to the end and a weight assigned at each point based on four factors: a measure of noisiness, a measure of curvature, contour length, and the gradient magnitude. The point with the largest average weight is chosen. After searching from the initial point in one direction, a similar search is conducted in the oppositedirection unless a closed contour has been formed. In the algorithm for multiple scales the search for a contour proceeds as for the single scale, using the largest scale, until a best partial contour at that scale has been found. Then the next finer scale is chosen and the neighborhood around the end points of the contour are examined to determine possible edge points in a direction similar to the end point of the contour. The original algorithm is then followed for each of the points satisfying the above condition, and the best is chosen as an extension to the original edge. Further, in order to determine the size neighborhood that should be searched when attempting to pick up an edge at a smaller scale, a theoretical analysis of the movement of idealized edges is performed. This analysis examines two adjacent step edges having the same parity (a staircase) and opposite parity (a pulse). It is determined that maximum movement for both cases is σ, where σ is the standard deviation of the Gaussian used. This maximum movement occurs for the staircase when the two nearby edges have the same step size and are at a distance of 2σ apart. However, for edges closer or farther away, maximum movement decreases rapidly. For a pulse, maximum movement occurs when the two edges have the same step size and are very close together. Again the movement decreases rapidly as the edges become farther apart. Movement also decreases in both cases when the relative strengths of the two edges are not equal.

使用多个尺度的边缘轮廓
讨论了一种利用高斯算子梯度检测的边缘点集寻找一条单优路径的算法。首先提出了一种单尺度轮廓检测算法,然后对该算法进行了扩展,利用多尺度对弱边缘进行了改进检测。可能的边缘点集被放置在优先队列中,其中最大的边缘点位于顶部。从队列中检索尚未在轮廓上的最强边缘点。首先检查计算方向上的点,然后检查其两侧相邻方向上的点。每个分支被跟踪到最后,并根据四个因素在每个点分配一个权重:噪声度量、曲率度量、轮廓长度和梯度大小。选择平均权值最大的点。从一个方向的初始点开始搜索后,除非已形成闭合轮廓,否则在相反方向进行类似的搜索。在多尺度算法中,对轮廓的搜索与对单尺度的搜索一样,使用最大的尺度,直到在该尺度上找到最佳的部分轮廓。然后选择下一个更精细的尺度,并检查轮廓端点周围的邻域,以确定与轮廓端点方向相似的可能边缘点。然后对满足上述条件的每个点按照原算法进行,并选择最优点作为原边缘的延伸。此外,为了确定在尝试以较小的尺度拾取边缘时应该搜索的邻域大小,对理想边缘的运动进行了理论分析。这个分析检查了两个相邻的阶边,它们具有相同的宇称(阶梯)和相反的宇称(脉冲)。确定两种情况下的最大移动是σ,其中σ是所使用的高斯的标准差。当相邻的两条边具有相同的步长,且距离为2σ时,楼梯就会出现最大的移动。然而,对于较近或较远的边缘,最大移动会迅速减少。对于脉冲,当两个边缘具有相同的步长并且非常靠近时,会发生最大的运动。同样,随着边缘之间的距离越来越远,运动也迅速减少。当两条边的相对强度不相等时,运动也会减少。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信