引入影像重构的边缘检测算子性能评估方法

Performance Evaluation of Edge Detectors Using Image Reconstruction

摘要: 首先,利用待评估边缘检测算子进行边缘检测；然后,根据检测边缘支撑区融合边缘位置信息与灰度信息，并利用改进的灰度反距离分区加权均值法实现影像重构；最后,设计反映原始影像与重构影像之间特征相似度的性能评估综合指标，完成边缘检测算子的性能评估。实验结果表明，该评估方法的结论与人工判读的结果相吻合，并且不需要先验参考信息，有一定的工程应用性。

Abstract: An objective method for performance evaluation of edge detectors was proposed based on image reconstruction. Firstly, binary edge maps were extracted by testing edge detectors. Then, combining the location information and brightness information in edge support region, image was reconstructed using enhance inverse of the distance for the weighted average of image pixel values in divisional regions. Finally, performance of edge detectors was evaluated using integrated index which represents feature similarity between reconstructed image and original image. The experimental results show that the proposed method is consonant with subjective judgment and has practical implementation.