一种山区图像边缘信息的分水岭变换提取方法  

Method of Extracting Image Edges in Mountain Areas Based on Watershed Transform

作　　者：魏振忠[1] 郗曼[1] 张广军[1] 何小妹[2] 

机构地区：[1]北京航空航天大学精密光机电一体化技术教育部重点实验室,北京100083 [2]中国航空工业第一集团公司北京长城计量研究所,北京100095

出　　处：《计算机测量与控制》2009年第7期1403-1406,共4页Computer Measurement &Control

基　　金：国家自然科学基金资助项目(50875014);教育部新世纪优秀人才支持计划项目(NCET-07-0043)

摘　　要：为了能够抑制引起过分割的根本原因——噪声和细密纹理,提出了一种改进的提取山区图的山脊线、山谷线的边缘信息的分水岭算法,该算法着重于解决传统分水岭的过分割问题;该算法首先对梯度图进行一种改进的多尺度梯度算子处理,以减小图像噪声对分水岭的过度分割的影响;然后利用基于标记符控制的最小强制技术去除由噪声和不必要的图像细节形成的谷底,使修正后的梯度图像只对应包含区域极大值的标记图像,改进后的方法能准确地提取出所需要的关键边缘信息。In order to eliminate over segmentation caused by the noise and high--texture as the major reason, an improved algorithm based on watershed transform for abstracting ridge and hollow edge of mountain images is presented, the algorithm focus on solving the over segmentation problem in traditional watershed transform. Firstly this algorithm uses a improved multi--scale algorithm to deal with the gradient of the original image, which can reduce the image noise＇s influence on over segmentation of watershed transform, and then add the improved minima imposition technology with marker controlled segmentation to the traditional watershed transform for eliminate ＇ bottoms＇ caused by noise and unnecessary image details. The processed gradient image corresponds to the marked--image having the regional maximum value, the improved algorithm can. precisely extract the needed and important edge information in image.

关 键 词：边缘提取 分水岭变换 多尺度梯度算子 强制最小技术 标记控制 

分 类 号：TP274.5[自动化与计算机技术—检测技术与自动化装置]