Aiming at solving the problem of imaging system degradation caused by the shortage of sensor Nyquist frequency and the point spread effect of optical system, we propose a modified super-resolution enhancement method for multiple input images in the frequency domain framework. In image registration, we build the phase difference model by considering the impact of spectrum aliasing and interpolation transformations, so the sub-pixel information obtained from the model is more accurate. Based on the simultaneous autoregressive prior, the joint Gaussian distribution model combined with the Bayes approach is built to restore the interpolation results. The experimental results demonstrate that the proposed algorithm is effective with low computational complexity.