<p>The last ten years witnessed the research process of the image segmentation to emerge and it was one of the multi-level thresholding algorithms as a regular procedure ready to yield results more frequently in the bi-level thresholding by virtue of simplicity, accuracy, efficiency, and fast convergence. The traditional methods would then have to be re-determined in situations where optimal multi-level thresholds should be determined to image segmentation. This paper thus presents an improved Search and Rescue Optimization Algorithm (SAR) to solve multi-level thresholding issues of fundus image segmentation. SAR is one of the well-known meta-heuristic algorithms which simulates human search and rescue reconnaissance and is extended to present the improved SAR algorithm. There is extensive testing that has been conducted in performance comparison of the efficiency in which the ISAR performs relative to solving multi-level thresholding problems in image segmentation. Comparative testing entails some of the algorithms mentioned above such as Harris Hawks Optimization (HHO), Sine Cosine Algorithm (SCA), Gravitational Search Algorithm (GSA), Arithmetic Optimization Algorithm (AOA), and the conventional SAR. Multi-level thresholding image segmentation with target functions such as fuzzy entropy, Kapur’s entropy, Tsallis entropy, and Otsu’s are employed to perform testing for multiple values of thresholds on test images. The results are compared to ensure Improved SAR preserves higher features and quality of segmented images over other methods. Mean Square Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Feature Similarity Index Metric (FSIM), and Normalized Correlation Coefficient (NCC) are employed as comparison criteria. The experiments clearly demonstrate the effectiveness of the SAR-enhanced algorithm compared to MSE, PSNR, FSIM, and NCC criteria in solving the problem of segmentation.</p>