Urban flooding poses a significant and growing threat to densely populated areas, creating an urgent need for accurate and rapid mapping tools to support disaster response and mitigation efforts. This study presents a comparative evaluation of object-based and pixel-based classification approaches for urban flood detection using Sentinel-1 Synthetic Aperture Radar (SAR) data. The proposed methodology incorporates key SAR-derived features, including backscatter-based indices, the Urban Flood Index (UFI), interferometric coherence, and texture metrics extracted using the Gray Level Co-occurrence Matrix (GLCM), to enhance flood mapping accuracy. This study also introduces an object-based multi-feature framework that integrates these SAR-derived indicators to improve urban flood delineation. Both approaches utilized Random Forest classifiers trained with high-resolution optical imagery as reference data. In the object-based method, multi-resolution segmentation was applied to group homogeneous image objects and extract aggregated feature values within each segment. This approach produced flood probability maps that closely aligned with visually interpreted reference data, accurately delineating inundated urban zones in Golestan Province, Iran. Conversely, the pixel-based method preserved fine spatial detail but showed increased sensitivity to speckle noise and misclassification, particularly in heterogeneous built-up areas. A quantitative assessment demonstrated that the object-based approach outperformed the pixel-based method, achieving an overall classification accuracy of 97.06% compared to 83.09%. These results highlight the value of incorporating spatial context and region-based analysis for effective flood detection. The integration of SAR-derived backscatter, coherence, and texture features within an object-based framework proves to be a reliable strategy for operational urban flood mapping.