A novel sparse array synthesis method for non-uniform planar arrays is proposed, which belongs to compressive sensing (CS)-based systhesis. Particularly, we propose an off-grid refinement technique to simultaneously optimize the antenna element positions and excitations with a low complexity, in response to the antenna position optimization problem that is difficult for standard CS. More importantly, we take into account the minimum inter-element spacing constraint for ensuring the physically realizable solution. Specifically, the off-grid Orthogonal Match Pursuit (OMP) algorithm is first proposed with low complexity and then off-grid Look Ahead Orthogonal Match Pursuit (LAOMP) is designed with better synthesis performance but higher complexity. In addition, simulation results have shown the proposed schemes have more advantages in computational complexity and synthesis performances compared with the related method.