Image super resolution (SR) of images is necessary for many computer vision and image processing applications. Deep learning research is still focused on examining the extraordinary resolution of infrared (IR) or thermal images. With the help of image super resolution technology, the drawback of infrared detectors with unnecessarily large pixel sizes has been successfully addressed, as the growing need for high resolution infrared image data. This work presents a single-image super resolution method for both visible and infrared images. It combines bicubic interpolation with deep learning. Low resolution images are compared to sample results based on original, higher resolution images by first undergoing bicubic interpolation and then using the deep learning method.