一种基于连通区域的轮廓提取方法

摘要

Contour extraction is one of the fundamental steps in computer vision. The correctness and reliability of its results affect directly the comprehension of machine vision systems for the objective world. There are many contour algorithms at present, and each has its own characteristics and shortcomings. Therefore, this paper gives an algorithm based on the connected regions and the threshold to not only extract the contour of the object but also erase the noise at the same step in image processing. Furthermore, the object can be located according to the contour. As shown in the experiments, the noises are erased clearly when the area of each noise is less than that of the object, and especially a single-pixel- width contour of the object can be presented which is also non-interval and non-cross, and not distorted from the object.%轮廓提取在许多智能视觉系统中被认为是非常重要的过程,其结果的正确性和可靠性直接影响到机器视觉系统对客观世界的理解.而现有诸多边缘检测的方法都存在着各自的局限性和不足之处,为此本文提出一种利用最佳阈值分割和基于连通区域面积阈值化的实现算法,可以同时实现噪声消除与轮廓提取,并据此定位图像中的物体目标.实验结果显示,只要噪声面积没有超过物体面积,应用该算法不仅可以完全消除噪声,而且能得到连续的无交叉的单像素宽度的物体轮廓,且轮廓不变形.