Improved Mural Image Segmentation Algorithm Using Northern Goshawk Optimization

Researchers at Hanyang University have developed a novel algorithm for restoring and protecting murals by leveraging optimization techniques. The proposed algorithm, OPBNGO, integrates the Northern Goshawk Optimization (NGO) algorithm with three learning strategies to enhance its global search ability, balance exploration and development phases, and improve development performance. The OPBNGO algorithm demonstrates promising results in image segmentation tasks, achieving a winning rate of over 96.87% in terms of fitness function value and over 93.75% in FSIM, SSIM, and PSNR metrics.

Key Takeaways:

• The OPBNGO algorithm is an improved Northern Goshawk Optimization algorithm designed for mural image segmentation.
• The proposed algorithm integrates three learning strategies: off-center learning, partitioned learning, and Bernstein-weighted learning.
• OPBNGO achieves a winning rate of over 96.87% in terms of fitness function value and over 93.75% in FSIM, SSIM, and PSNR metrics.
• The algorithm demonstrates promising results in image segmentation tasks for eight mural images.
• OPBNGO's development is attributed to researchers Jianfeng Wang, Zuowen Bao, and Hao Dong from Hanyang University's College of Design.
• The study highlights the potential of biomimetics in restoring and protecting cultural heritage sites.

Statistics:

• OPBNGO achieves a winning rate of over 96.87% in terms of fitness function value.
• The algorithm achieves a winning rate of over 93.75% in FSIM, SSIM, and PSNR metrics.
• The study involves eight mural images for image segmentation tasks.
• The proposed algorithm demonstrates improved performance compared to existing optimization-based image segmentation methods.

Sources:

• Biomimetics, "An Improved Northern Goshawk Optimization Algorithm for Mural Image Segmentation" (2025, 10(6), 373).
• MDPI AG, publisher of Biomimetics.
• https://doi-org.sdpl.idm.oclc.org/10.3390/biomimetics10060373