Abstract: Edge detection is an important step in image processing. The results of edge detection effect directly the results of image processing. But there is lacking a standard numerical evaluation measure for the results of edge detection. A method to evaluate the edge connectivity for the results of edge detectors is brought forward. This paper regards the edge connectivity as an evaluation index of the results of image edge extraction, and evalutes the edge connectivity by calculating the average value of the product of the area of convex?hull and the length of edge segments. This paper designs multiple tests to assess the performance of the evaluation algorithm on one image by different edge detection algorithms and different parameters. The experiments prove that the algorithm of evaluating the edge connectivity is effective and fast, the evaluation result is conform to human visual system. This method has some application value for the some high level of image processing and automatic processing.

Key words: image processing, edge detection, edge detection evaluation, numerical evaluation, edge connectivity

摘要： 边缘检测是图像处理的一个重要环节，边缘检测效果的好坏直接决定图像处理结果的好坏，但对于边缘检测结果缺乏一个标准的数值化的评价方式，因此提出了一种对边缘检测结果的边缘连续性量化评价方法。以边缘的连续性作为图片边缘提取效果的评价指标，并且使用边缘段凸包面积与边缘段长度的乘积的平均值来数值化评价边缘连续性。设计了多种实验对同一图片采用不同的边缘检测算法在不同边缘检测参数下进行检测，并与所提算法进行评价对比。实验证明该算法能数值化地快速、有效地评价图片边缘检测的效果好坏，评价结果符合人的视觉感知特征，对于高层次的图像处理与自动化图像处理环节具有较好的应用价值。

关键词: 图像处理, 边缘检测, 边缘检测评价, 量化评价, 边缘连续性