Citation: Al Darwich, R.; Babout, L.;
Strzecha, K. An Edge Detection
Method Based on Local Gradient
Estimation: Application to
High-Temperature Metallic Droplet
Images. Appl. Sci. 2022, 12, 6976.
https://doi.org/10.3390/
app12146976
Academic Editor: Antonio Fernández
Received: 27 April 2022
Accepted: 1 July 2022
Published: 9 July 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
An Edge Detection Method Based on Local Gradient Estimation:
Application to High-Temperature Metallic Droplet Images
Ranya Al Darwich *, Laurent Babout * and Krzysztof Strzecha *
Institute of Applied Computer Science, Łód´z University of Technology, 90-537 Łód´z, Poland
* Correspondence: raldarwich@kis.p.lodz.pl (R.A.D.); laurent.babout@p.lodz.pl (L.B.);
strzecha@kis.p.lodz.pl (K.S.)
Abstract:
Edge detection is a fundamental step in many computer vision systems, particularly in
image segmentation and feature detection. There are a lot of algorithms for detecting edges of objects
in images. This paper proposes a method based on local gradient estimation to detect metallic droplet
image edges and compare the results to a contour line obtained from the active contour model of
the same images, and to results from crowdsourcing to identify droplet edges at specific points. The
studied images were taken at high temperatures, which makes the segmentation process particularly
difficult. The comparison between the three methods shows that the proposed method is more
accurate than the active contour method, especially at the point of contact between the droplet and
the base. It is also shown that the reliability of the data from the crowdsourcing is as good as the edge
points obtained from the local gradient estimation method.
Keywords: edge detection; active contour; gradient-based edge detection; crowdsourcing
1. Introduction
Edges have been defined as significant local variations in image intensity and are
considered an essential feature of image analysis. Given the importance of edge detection
in image processing, image analysis, image pattern recognition and computer vision
techniques, edge detection has continued to be an effective research area and a fundamental
step in retrieving information from digital images [
1
]. There are many edge detection
methods that have been developed by several researchers [
2
–
4
]. However, there are some
factors that affect the performance of the edge detection process, especially in the case of
images with low-intensity differences between adjacent regions. These differences lead
to blurring of the boundaries between these regions. Blurred boundaries can affect the
accuracy of the edges delineation and make edge detection a challenging task [
5
]. This
paper addresses the problem of detecting edges in images of hot specimens of molten
copper that emit thermal radiation since the illumination of the specimen’s boundaries can
be considered an additional factor affecting image quality [6].
The aim of the research in this paper is to propose an edge detection method based on
the local gradient estimation to define the edges of a metallic droplet images obtained from
the THERMO-WET system for high temperature measurements of surface properties and
compare the edges to the contour line obtained from the active contour model of the same
images [7], and crowdsourcing-based detection [8].
The studied images show that the edges are noticeably blurred due to the thermal
effects. The edge detection method presented in this paper prove to give better results
compared to the classic active contour method [9].
2. Materials and Methods
The studied images were obtained from the THERMO-WET system, as the first in
the world, which enabled automated measurement in a wide temperature range (up to
Appl. Sci. 2022, 12, 6976. https://doi.org/10.3390/app12146976 https://www.mdpi.com/journal/applsci
