Abstract In recent years, the effective processing of high-resolution color satellite images obtained through remote sensing technologies has become a critical requirement in key applications such as environmental monitoring, urban planning, and disaster management. Color satellite imagery offers rich information, enabling more detailed analysis of land use, vegetation cover, and other surface characteristics. In this context, multi-level image thresholding techniques are widely employed to enhance segmentation quality; however, achieving both high accuracy and low computational cost in complex scenes remains a significant challenge. To address these challenges, this study proposes the RSA-HGSO algorithm, an adaptive hybrid structure that combines the Reptile Search Algorithm (RSA) and Henry Gas Solubility Optimization (HGSO). By integrating RSA’s global exploration capabilities with HGSO’s local exploitation strengths, RSA-HGSO ensures both solution diversity and fast, stable convergence. Experimental analyses conducted on high-resolution color satellite images from the NASA Visible Earth dataset demonstrate the algorithm’s efficient performance. With average values of PSNR = 24.59 dB, SSIM = 0.9088, FSIM = 0.9233, and NCC = 0.9663, RSA–HGSO outperforms comparative methods in terms of both accuracy and structural integrity. Furthermore, correlation analyses indicate that the original pixel intensity ranking is preserved at a rate of 91% after segmentation (Pearson = 0.9557, Spearman = 0.9549), and neighborhood relationships are largely maintained (Pearson = 0.8042, Spearman = 0.8051). Ablation studies further confirm that RSA–HGSO not only integrates the performance of its component algorithms but also complements exploration and exploitation processes in a synergistic manner, resulting in more balanced and robust outcomes. The findings suggest that the RSA–HGSO algorithm offers a balanced, scalable, and computationally efficient solution to the problem of color multi-level satellite image thresholding and presents a practical method applicable to real-world tasks such as disaster management, agricultural monitoring, and urban planning.
Şahin et al. (2026) studied this question.