This study investigates the effectiveness of high-resolution optical (IKONOS) and radar (TerraSAR-X) satellite imagery for automatic road extraction in urban areas using texture features integrated into neural network algorithms. The authors compare the performance of these image sources by applying statistical texture measures—such as entropy, contrast, and homogeneity—and feeding them into a back-propagation neural network. Accuracy assessments using metrics like RCC, BCC, and RMSE reveal that each image type has unique advantages and limitations: optical imagery performs better in vegetated regions, while radar imagery is more robust in areas with dense urban infrastructure and narrow roads. The findings emphasize that combining radar and optical data can improve road detection accuracy and suggest a hybrid approach as a promising direction for future research in urban remote sensing applications.