Segmentation of thermographies from electronic systems by using the global-best brain storm optimization algorithm
No Thumbnail Available
Authors
El-Abd, Mohammed
Mousavirad, Seyed Jalaleddin
Nadimi-Shahraki, Mohammad H.
Navarro, Mario A.
Oliva, Diego
Ortega-Sanchez, Noé
Ramos-Michel, Alfonso
Issue Date
2023-12-01
Type
Journal Article
Language
Keywords
Alternative Title
Abstract
Segmentation is an important and basic task in image processing. Although no unique method is applicable to all types of images (as thermographies), multilevel thresholding is one of the most widely used techniques for this purpose. Multilevel thresholding segmentation has a major drawback that is to properly find the best configuration of thresholds. For that reason some metaheuristic algorithms are used to optimize the searching for the best thresholds. This paper proposes a combination of the minimum cross-entropy method and the Global-best brain storm optimization algorithm (GBSO), which improves the standard BSO to find the optimal solutions in complex search spaces. The GBSO uses a population of agents based on a global best and a re-initialization scheme that is triggered by the current state of its population. Here, the GBSO is used to find the best configuration of thresholds by optimizing the minimum cross entropy that is commonly using in image segmentation. Once the best thresholds are obtained they are applied over the images to extract only the regions of interest. For example, in the case of thermographies the parts with higher temperatures. To verify the performance of the proposed method it is firstly applied to classical reference images and after that over thermal images from electronic devices. The idea is to provide an alternative to segment thermographies that permits separating regions with higher temperatures. This could be used as a preprocessing step in a complex image processing system. The experimental result in terms of segmentation of electronic devices in thermographies provides evidence of the good performance of the GBSO. Different comparison with recent methods from the state-of-the-art were conducted where the GBSO obtains 1st place with the best values for the MCET. To validate the quality of segmentation they were used metrics as the peak signal-to-noise ratio (PSNR) where the GBSO is in the 4th rank of comparison, the structural similarity index (SSIM) and the feature similarity index (FSIM). For the FSIM and SSIM the GBSO in the 4th and 3rd rank, respectively.
Description
Citation
Publisher
Springer Netherlands
License
Journal
Volume
82
Issue
29