Improved Harris Hawk Optimization for Multilevel Thresholding Image Segmentation

Strategic Insights -

Researchers at Shandong University in China have developed a new approach to image segmentation using the Harris Hawks Optimization (HHO) algorithm. The team used logistic chaotic mapping to increase the diversity of solutions, introduced the Cauchy mutation operator and opposition-based learning (OBL) strategy to perturb the optimal solution position, and incorporated a Gaussian mutation to improve the global search ability. The proposed algorithm was compared to existing methods on six benchmark images and eight images from the Berkeley BSD500 datasets, resulting in a superior performance.

Key Takeaways:

  • The Harris Hawks Optimization (HHO) algorithm was modified to address issues of instability and imbalance in exploration and evolution capabilities.
  • The proposed method uses logistic chaotic mapping to increase population diversity and introduces the Cauchy mutation operator and opposition-based learning (OBL) strategy to perturb the optimal solution position.
  • The algorithm was compared to existing methods using eight metrics, including PSNR and SSIM, on six benchmark images and eight images from the Berkeley BSD500 datasets.
  • The experimental results showed that the proposed algorithm ranks first compared to other algorithms, with a Friedman mean rank of 1.6.
  • The proposed algorithm produces efficient and reliable results in terms of both consistency and accuracy.
  • The study demonstrated the effectiveness of the proposed approach on the CEC 2020 benchmark function test suite.
  • Financial supporters for this research include the National Key R&D Program of China and the Major Scientific and Technological Innovation Project of Shandong Province.
  • The research team includes Peng Yao, Xuwei Du, Qilin Wang, Xiang Liu, Mingwu Hao, Dongkai Chu, and Shuoshuo Qu from Shandong University.

Statistics:

  • The proposed algorithm was compared to 7 other algorithms on the CEC 2020 benchmark function test suite.
  • The experimental results were verified using Friedman test statistics.
  • The proposed algorithm produced an average Friedman mean rank of 1.6 compared to other algorithms.
  • The study used 8 metrics to evaluate the performance of the proposed algorithm, including PSNR and SSIM.
  • The algorithm was tested on 6 benchmark images and 8 images from the Berkeley BSD500 datasets.

Sources:

  • NewsRx. New Mathematics Findings Has Been Reported by Investigators at Shandong University (Improved Harris Hawk Optimization for Multilevel Thresholding Image Segmentation). Life Science Weekly. October 21, 2025; p 3331.
  • Improved Harris Hawk Optimization for Multilevel Thresholding Image Segmentation. Cluster Computing, 2025;28(13).
  • Springer. Cluster Computing. www.springer.com; Cluster Computing. www.springerlink.com/content/1386-7857/