Abstract
This article presents a comparative study of the super image resolution method using Deep Image Priors, comparing it with other advanced methods. The method uses the architecture of a deep neural network as a fixed a priori factor for the image restoration problem, and an optimization algorithm is applied to find the optimal set of weights for the network. The evaluation of the method's performance was conducted using the SET 5 dataset using the PSNR metric and compared with the RealESRGAN method. This study highlights the effectiveness of the Deep Image Prior method for super-resolution images and provides insights into its potential for further improvement. The Deep Image Priors method shows impressive results in super-resolution images, and its ability to use deep neural networks as a priori knowledge opens the way for the development of more efficient methods in the future. This study also highlights the importance of optimizing neural network weights to achieve the best results in image restoration tasks. Through benchmarking with RealESRGAN, it's exciting to witness the potential impact of Deep Image Priors in the field of image processing and computer vision. It seems to have the capacity to enhance the quality of super-resolution images significantly. It'll be interesting to see how we can further improve its effectiveness in the future.