Abstract: Prior information and constraints are of importance in super-resolution image reconstruction which is always an ill-posed problem. In this paper, first, a degradation model for passive millimeter wave imaging is established and the point spread function （PSF） of the model is estimated. Then, by introducing radiation characteristics of typical objects and scene as a-prior knowledge, a method combining Projection Onto Convex Set （POCS） with Maximum A Posterior （MAP） is proposed for optimum image reconstruction. Experiment restdts show that the method proposed in this paper has good convergence and can preserve the textures well while reducing noise in the image.