Recent developments in remote sensing, particularly deep learning-based Single Image Super-Resolution (SISR) techniques, have improved the spatial detail of satellite imagery, enabling more accurate mapping of land cover types such as mangrove forests. This study evaluates and compares the effectiveness of two datasets—standard Sentinel-2 imagery (10 m resolution) and super-resolved Sentinel-2 Deep Resolution 3 (S2DR3) data—in mapping mangrove forests across the Qeshm, Sirik, and Gwatar regions. Two widely used classification algorithms, Random Forest (RF) and Support Vector Machine (SVM), were applied to each dataset. The results showed that the RF algorithm applied to S2DR3 data achieved the highest classification accuracy (96%) for mangrove detection, outperforming RF–Sentinel-2 (86%), SVM–S2DR3 (90%), and SVM–Sentinel-2 (81%). In terms of area estimation, RF–S2DR3 mapped 8,147 ha of mangroves, while RF–Sentinel-2, SVM–S2DR3, and SVM–Sentinel-2 yielded 8,023, 9,227, and 9,984 ha, respectively. These results demonstrate that the higher spatial resolution of S2DR3 enhances the detection of smaller, fragmented mangrove patches, particularly along coastal fringes. Overall, the findings confirm the superiority of both S2DR3 imagery and the RF algorithm in accurately classifying and quantifying mangrove cover. This approach holds promise for improving the precision of ecosystem monitoring and supporting more effective conservation strategies in sensitive coastal environments.