In recent years, algorithms have been developed to decompose images into structural and textural components. Radar images, known for all-weather capabilities, prove increasingly valuable in equatorial regions with quasi-permanent cloud cover. As the African population undergoes rapid demographic growth, it becomes imperative to establish automatic mapping methods for radar images. This facilitates timely urban evolution studies within territorial jurisdictions. This paper presented an approach to delineate urban boundaries in synthetic aperture radar (SAR) images from European Remote Sensing Satellite (ERS)-1, focusing on Douala, Cameroon. Our method emphasizes detecting key features, including road networks, urban structures, inhabited areas, and wetlands, considering the unique granularity of SAR imagery. The decomposition technique was applied to road network detection in aerial or satellite imagery. The algorithmic chain employs a combination of textural image analysis and mathematical morphology, utilizing set theory for road network extraction from radar images and aerial photography. Additionally, linear structures are detected through mathematical morphology, followed by the refinement of extracted roads using statistical active contours. Applying wavelet transforms, filtering by rectangularity measure, histogram analysis, Canny contour detection, morphological operators, and watershed transformation demonstrated precise road detection, effectively distinguishing them from other objects. The algorithm proves efficient across urban, rural, and peri-urban contexts. For inhabited areas and wetlands, the algorithm adapts to higher gray levels associated with rooftops and accurately detects extremely low gray levels indicative of water bodies.