Microwave coincidence imaging (MCI) is a novel staring imaging technique with high resolution in azimuth. In MCI, the continuous imaging area is discretized into fine grids and the target-scattering centers are assumed to be exactly located at the centers of prediscretized grids. Recently, parametric methods are applied to MCI as target reconstruction algorithms with resolution enhancement and quality improvement. However, in practical applications, grid mismatch will severely degrade the imaging quality of parametric methods because the target-scattering centers will not totally locate at the grid centers no matter how fine the grids are. In this article, a reweighted-dynamic-grid-based MCI (RDG-MCI) method is proposed. In RDG-MCI, grids are evolving from coarse to dense iteratively rather than being fixed, and hence, off-grid errors can be eliminated gradually. Meanwhile, the reconstructed coefficients are used as weighting factors of grids in a form of weighting matrix in the next iteration and nonkey grids will be dropped out. Hence, the dynamic grids will be focused around the positions where target scatterers are most likely to exist. Furthermore, the matrix uncertain sparse Bayesian learning (MUSBL) algorithm is adopted to eliminate the residual off-grid errors. Finally, a preferable imaging result can be obtained based on the updated nonuniform grids. Also, the theoretical expected Cramér–Rao bound (ECRB) is also derived to evaluate the performance of the proposed method. The effectiveness of the proposed method, along with the super-resolution ability of MCI, is verified by simulations and outdoor experiments.