<p>Semantic segmentation of remote sensing images is a fundamental task for critical applications like land cover mapping and urban planning. However, the inherent complexity, large scale variations, and intricate spatial patterns of remote sensing data pose significant challenges to achieving high segmentation accuracy. To address these issues, we propose the Independent Dual-Branch Network (IDBNet). Our architecture features a dual-branch encoder consisting of a Mamba-based branch and a convolutional neural network (CNN) branch, which operate independently to extract complementary global context and fine-grained local details, respectively. A decoder built upon dynamic sampling principles progressively restores spatial resolution with high fidelity. A key innovation is our Triple Feature Completion Module (TFCM), designed to effectively integrate these diverse feature streams. The TFCM refines and fuses global representations from the Mamba branch, local details from the CNN branch, and, crucially, intermediate features recovered during the decoding process. Experimental results on the ISPRS Vaihingen and LoveDA datasets demonstrate that IDBNet achieves state-of-the-art performance, validating the efficacy of our approach in handling challenging remote sensing scenes.</p>