The image acquisition process in the field of magnetic resonance imaging (MRI) does not always provide high resolution results that may be useful for a clinical analysis. Super-resolution (SR) techniques manage to increase the image resolution, being especially effective those based on examples that determine a correspondence between patterns of low resolution and high resolution. Deep learning neural networks have been applied in recent years to estimate this association with very competitive results. In this work, the starting point is a convolutional neuronal network to which a regularly spaced shifting mechanism over the input image is applied, with the aim of substantially improving the quality of the resulting image. This hybrid proposal has been compared with several SR techniques using the peak signal-to-noise ratio, structural similarity index and Bhattacharyya coefficient metrics. The results obtained on different MR images show a considerable improvement both in the restored image and in the residual image without an excessive increase in computing time.