摘要： 针对边缘检测算法的性能评价，提出了一种在真实场景下基于统计、自适应强的评价基准图计算方法。定义了算法参数的相关系数，研究了算法参数相关模型。根据建立的算法参数相关系数得到单算法的预选基准图，对不同算法预选基准图进行基于置信度的图像融合得到可用于边缘检测算法性能评价的基准图。该计算基准图的方法有助于实现边缘检测算法性能评价的自动化，实验结果验证了该方法的有效性及实用性。

关键词: 置信度, 相关性, 图像融合, 边缘检测性能评价, 基准图

Abstract: This paper proposes a new method for calculating the ground truth image by image fusion, and designs the calculation approach for the processing. Using the correlation coefficient between the different detection results, reference ground truth(RGT) images for each detector is obtained. The ground truth (GT) images are calculated by image fusion based on confidence measures, using the RGTs. The method suggests a general tool to assist in practical evaluation of parametric edge detectors where an automatic process is required.

Key words: confidence measures, correlation, image fusion, edge detection evaluation, ground-truth image

中图分类号: 

• TP391