MRI is sensitive to motion caused by patient movement. It may cause severe degradation of image quality. We develop an efficient retrospective deep learning method called stacked U-Nets with self-assisted priors to reduce the rigid motion artifacts in MRI. The proposed work exploits the usage of additional knowledge priors from the corrupted images themselves without the need for additional contrast data. We further design a refinement stacked U-Nets that facilitates preserving of the image spatial details and hence improves the pixel-to-pixel dependency. The experimental results prove the feasibility of self-assisted priors since it does not require any further data scans.