Scholarlink

<p>The skin, our body’s primary defence, is susceptible to numerous health issues. Early detection and accurate diagnosis of such conditions rely heavily on medical image processing. Multi-level thresholding image segmentation is a leading technique in this field, requiring an efficient optimizer to achieve optimal results. The Single Candidate Optimizer (SCO), a metaheuristic algorithm, has shown promise in optimization but is limited by its reliance on a single solution, leading to restricted diversification and difficulty escaping local optima. To overcome these challenges, we introduce a population-based version of SCO, termed PSCO. This approach enhances SCO by integrating an influencing strategy to guide the population’s swarming behaviour, and the inverse incomplete gamma function to deepen the intensification phase. In addition, a soft strategy is developed to balance diversification and intensification, ensuring a smooth transition from global to local search. PSCO was rigorously tested on the CEC 2022 benchmark suite across 10 and 20 dimensions, consistently outperforming SCO. Furthermore, the effectiveness of PSCO is validated in the context of multi-level thresholding image segmentation, applied to two datasets: BSD500 for benchmarking and ISIC for medical imaging. The algorithm’s performance is evaluated using Otsu’s and Kapur’s fitness functions across different thresholds. Additionally, its effectiveness is rigorously assessed by comparing it against ten state-of-the-art solutions, focusing on key metrics such as Otsu’s and Kapur’s objective values, PSNR, SSIM, FSIM, QILV, HPSI, UIQI, and CPU time. The results demonstrate significant improvements, highlighting PSCO’s potential in general optimization tasks and specialized applications such as medical image segmentation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A population-based variant of the Single Candidate Optimizer for global optimization and multi-level thresholding problem in medical image segmentation

  • 期刊论文

  • 发表日期：

    2025年7月8日

查看全文
  • Abdelhadi Limane,
  • Farouq Zitouni,
  • Saad Harous,
  • Rihab Lakbichi,
  • Aridj Ferhat

摘要

The skin, our body’s primary defence, is susceptible to numerous health issues. Early detection and accurate diagnosis of such conditions rely heavily on medical image processing. Multi-level thresholding image segmentation is a leading technique in this field, requiring an efficient optimizer to achieve optimal results. The Single Candidate Optimizer (SCO), a metaheuristic algorithm, has shown promise in optimization but is limited by its reliance on a single solution, leading to restricted diversification and difficulty escaping local optima. To overcome these challenges, we introduce a population-based version of SCO, termed PSCO. This approach enhances SCO by integrating an influencing strategy to guide the population’s swarming behaviour, and the inverse incomplete gamma function to deepen the intensification phase. In addition, a soft strategy is developed to balance diversification and intensification, ensuring a smooth transition from global to local search. PSCO was rigorously tested on the CEC 2022 benchmark suite across 10 and 20 dimensions, consistently outperforming SCO. Furthermore, the effectiveness of PSCO is validated in the context of multi-level thresholding image segmentation, applied to two datasets: BSD500 for benchmarking and ISIC for medical imaging. The algorithm’s performance is evaluated using Otsu’s and Kapur’s fitness functions across different thresholds. Additionally, its effectiveness is rigorously assessed by comparing it against ten state-of-the-art solutions, focusing on key metrics such as Otsu’s and Kapur’s objective values, PSNR, SSIM, FSIM, QILV, HPSI, UIQI, and CPU time. The results demonstrate significant improvements, highlighting PSCO’s potential in general optimization tasks and specialized applications such as medical image segmentation.