The task of multistatic radar imaging can be converted to the problem of jointly sparse signal recovery. In this paper, the algorithm named decentralized and collaborative subspace pursuit (DCSP) is utilized in multistatic radar systems to obtain a high-resolution image. By embedding collaboration among radar nodes and fusion strategy into each iteration of the standard subspace pursuit (SP) algorithm, DCSP is capable of providing satisfactory image even if some radar nodes suffer from relatively low signal-to-noise ratios (SNRs). Compared to the existing algorithms based on linear programming, DCSP has much lower computational complexity at the cost of increased communication overhead in the radar network.