An approach is presented for separating and compensating wideband signatures of radar targets. The signatures thus modified are processed into focused radar imagery. In this approach, the time- frequency representation of superimposed signatures is utilized to define selective binary filters that are enhanced and labeled using computer vision techniques. On filtering, the signature of each target is synthesized from its short-time Fourier transform representation. Following synthesis, the filtered signature is motion compensated in the frequency domain. Radar images can then be generated for scenarios in which multiple moving targets cannot be imaged otherwise. Data collected with an actual stepped frequency radar system is processed to demonstrate that combined signatures of aircraft are effectively separated, compensated and transformed into individual inverse synthetic aperture radar (ISAR) images.