Image Super-Resolution Using Auto Encoders

Image super-resolution (SR) remains a crucial problem in computer vision to obtain high resolution images from low-resolution ones. In this paper, a novel approach towards SR is presented with the help of auto-encoders, a specific kind of neural network whose task is to learn compressed representations of the high resolution images. In the context of the proposed model, there is an encoder, which, alternately, compresses the input and a decoder which reconstruct the output is proposed as capturing as well as reproducing all the details of an image. Experimental results show that the proposed method is one of the most efficient and accurate ones on benchmark datasets in terms of better details preservation and artefact reduction besides reducing processing time compared to the conventional methods. This makes this approach to be flexible and efficient to be used in medical imaging, satellite and digital photography making it an effective solution to improving image quality and resolution.