Analysis of Edge Detectors based on Performance Evaluation Metrics
A joint comparison of the outcomes of edge detection techniques provides a more thorough knowledge of the capabilities of the algorithms to identify image edges. Edge detection methods are classified as conventional edge detection, soft computing edge detection and deep learning edge detection strategies. Detection efficiency is evaluated by means of a comprehensive analysis that includes F1-score and Intersection over Union (IoU). Deep learning-based models have proven to be more efficient with regard to F1-scores due to better feature extraction. High PSNR indicates good agreement between detected boundaries and actual objects. In the course of conducted experiments, Clip-MobileNetV2-Unet has shown better results concerning Recall, Precision, F1-Score/mDC, Accuracy, PSNR (dB), and mIoU. All these metrics make up a complex evaluation strategy. From this study, it is found that deep learning algorithms perform better in terms of accuracy and reliability compared to other approaches, while traditional detectors are fast and suitable for simple real-time purposes.