一种基于种子生长的可控边缘检测法及评价方案  

A Seeds Growing Based Controllable Edge Detection Method and an Evaluation Protocol

作　　者：王捷[1] 莫灿林[1] 曹国惠[1] 

机构地区：[1]浙江理工大学机械与自动控制学院,杭州310018

出　　处：《浙江理工大学学报（自然科学版）》2010年第5期738-744,共7页Journal of Zhejiang Sci-Tech University(Natural Sciences)

基　　金：浙江省重点基金项目(Z6090572)

摘　　要：分析传统算子法的缺陷并进行有效的改进。提出一种基于种子生长的边缘检测方法,采用Laplace算子的一种改进形式作为模板,在卷积前预先截断小灰度差分,选择高卷积像素作为种子进行生长,并对生成的边缘进行后续处理,从而得到最终的二值边缘图像。实例的调控与分析表明,所提方法的检测效果具有良好的定位精度、边缘响应、连续性和可控细节。最后给出一种评价方案,对多种方法进行比较分析并证明该方法的优势。Analyzes the disadvantages of traditional operators and improves them efficaciously. Proposes a controllable detection method based on seeds growing, uses a modified form of laplacian as the template, blockes small grayscale differences before convoluting, adopts high convolution pixels as seeds and let them grow, takes subsequent process on the generated edge, thus achieves the final two-value edge image. Example＇s regulating and analyzing shows that the raised method has nice location precision, edge respondency, continuity and steerable details. In the end, provides an evaluation protocol, compares kinds of methods and proves the advantages of the mentioned method.

关 键 词：种子生长 可控细节 定位精度 连续性 评价方案 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]