Cooperative radar networks are a promising technology in various areas, such as vehicleto-infrastructure networks for automotive radar and radar remote sensing with UAVs. The use of widely distributed radar networks enables the detection of targets with complex scattering characteristics, as their coherent bistatic images are superior for forward scattering, and each monostatic image illuminates a scene from a different perspective. This work introduces a signal processing scheme that addresses two main challenges in this area: the coherent signal processing of uncoupled radar nodes and the self-localization of the nodes for radar image combination. A comprehensive signal model that incorporates time, frequency and phase incoherency is introduced. Based on this, an algorithm for constellation estimation, synchronization up to the carrier phase level, and multiperspective imaging is developed. The proposed approach is experimentally verified using commercially available 77 GHz single-input/multiple-output radar nodes. The measurements for different radar constellations and various target scenes show a self-localization accuracy below 6 cm in range and below 2.5 • for the incident angles. The resulting images of various scenes clearly indicate an information gain compared to single monostatic images due to the combination of bistatic and multiperspective monostatic images.