<p>Image segmentation is fundamental in various domains, including medical imaging, satellite analysis, and microscopic studies. The watershed transform is a widely adopted method for delineating object boundaries; however, its susceptibility to noise and over-segmentation limits its effectiveness. This study proposes two enhanced strategies for marker generation in marker-controlled watershed segmentation. The first, Dual-Channel Morphological and Gradient Spectral (DMGS), integrates morphological and spectral features to generate more informative markers and reduce over-segmentation. Building upon this, the second method, Hybrid Marker Extraction by Gradient and Spectral features (HMEGS), refines the process by eliminating redundant morphological markers and introducing an adaptive thresholding mechanism based on gradient analysis. HMEGS further enhances boundary precision and segmentation consistency by preserving dominant edge structures and improving region homogeneity. Experimental evaluations on the BSDS500 dataset demonstrate that HMEGS achieves a 0.17% reduction in Variation of Information (VI), a 1.4% increase in Covering (CV), and a 0.5% improvement in the Probabilistic Rand Index (PRI) compared to state-of-the-art methods. These results highlight the robustness and accuracy of the proposed framework across diverse imaging scenarios.</p>