X-Ray micro-computed tomography (micro-CT) is a standard method to perform three-dimensional analysis of the internal structure of a rock sample. 3D X-Ray microscopes, such as those from the XRadia Versa family, provide images of high resolution and contrast. Medical scanning machines can also be used for scanning rock samples to reduce operational cost and time, but they generally provide poorer spatial resolution and contrast compared to 3D X-Ray microscopes. Recent success in implementing deep learning algorithms to enhance image quality demonstrated that, in some cases, the application of convolutional neural network (CNN) models might significantly enhance the resolution of the micro-CT images. In this research, a super-resolution technique employing the U-Net 3D CNN architecture is applied to enhance the resolution of granodiorite rock sample images obtained by two different 3D scanning machines. The high-resolution dataset was obtained using the XRadia Versa XRM-500 microscope. It contained images with nominal resolutions of 10.3 and 5 microns. The low-resolution scanning was performed using a Scanco medical µCT 50 machine, and the images from this dataset had a nominal resolution of 10.3 microns. Several models were created to enhance the quality of the low-resolution images, and the results were analysed. It was observed that super-resolution processing could significantly improve the low-resolution micro-CT image quality and suppress noise that appeared on medical images. The results presented in this study are of particular interest and value to geoscientists that use medical scanners to study the structure of rock samples at large scale.