摘要
图像处理尤其是区域分析中,区域的边界是非常重要的信息。提出了一种基于优先搜索方向的边界跟踪方法。该方法是以区域为基本操作单元,以当前边界点和上一边界点所确定的方向的外法线方向为起始搜索方向,顺时针方向搜索下一个边界点,当下一个边界点与该区域第一个边界点重合时,该区域搜索完毕,并开始对下一个区域进行跟踪。利用这种方法跟踪出来的区域边界,不但能够很好的满足对边界一个像素的宽度、同一个区域的边界连通封闭、准确定位在区域外围轮廓上的要求,而且能够完整准确地保持目标区域外边界的自身原貌和特点,以及内、外边界彼此独立,可独立成图用。由于这种算法在跟踪过程中充分考虑已知边界点对后续边界点的影响,使得边界点保持了良好的相关关系,可以根据实际应用的要求,在跟踪过程中直接实现边界的矢量化。
In digital image processing, especially in region analying, boundaries and edges are very important information. In this work, the author presents a boundary tracking method based on the direction priority for searching. This method, whose the operation unit is region, takes the normal direction of the direction which is built by current boundary point and previous boundary point as the first direction to clockwise detect next boundary point in 4-neighbour or 8-neighbour of the current boundary point. When the tracking probe meets the first boundary point of the region, the tracking of this region is finished and begins the next tracking of next region. The boundary extracted with this method from image, not only retains the basic expectation of a boundary as a pixel width, connectivity and exactly locating on the contour of a region, present the periphery characteristic of a region, but also the boundary and inner-edge, which are caused by the holes in the region, can be easily separated from each other and either can be output as a thesis map for use. As the method taking well notice of the action which previous boundary points impact upon the posteriors, boundary points can keep their correlations. Because of these foundations, grid-formatted boundaries can be easily translated into vectors during the course of tracking following the requirement of application in practice.
出处
《遥感技术与应用》
CSCD
2004年第3期209-213,共5页
Remote Sensing Technology and Application
基金
上海市科学技术基金2002年专项
"十五"国家科技攻关项目和国家863计划共同资助。
关键词
边界跟踪
边缘检测
区域分割
图像处理
Boundary tracking, Edge detection, Image segmentation, Image processing
