Enhanced Holistically Nested Edge Detection (eHED) Algorithm: A Reliable Edge Detection in Unconstrained Scenarios

Understanding scenes, interpreting images, and identifying objects are all areas of computer vision that depend on edge detection. Many edge detection techniques, including traditional edge detection algorithms, have already been developed. The holistically nested edge detection (HED) system can accurately detect edges even under challenging circumstances, but it requires further improvements in its edge detection method. To address the constraints of the HED algorithm, this research introduces a novel enhanced HED (eHED) method. Integrating spatial attention (SA) into the HED method, eHED improves the ability to find edges, especially in complex scenarios. To improve the accuracy of edge detection, eHED focuses on informative parts of the image by successfully using spatial attention. This results in the exclusion of noise and features that are irrelevant to the current scenario. Extensive experiments are carried out to evaluate the effectiveness of the eHED algorithm on standard datasets including the ALPR dataset and PKU (License Plate Detection) dataset. The eHED algorithm, a substantial progress in edge detection technology, enhances performance and achieves an accuracy rate of 84% when deployed to the given dataset.