Magnetic resonance imaging (MRI) is a common way to diagnose related diseases. However, the magnetic resonance (MR) images are easily defected by motion artifacts in their acquisition process, which affects the clinicians' diagnosis. In order to correct the motion artifacts of MR images, we propose a convolutional neural network (CNN)-based method to solve the problem. Our method achieves a mean peak signal-to-noise ratio (PSNR) of (35.212±3.321) dB and a mean structural similarity (SSIM) of 0.974 ± 0.015 on the test set, which are better than those of the comparison methods.