ABSTRACT
The process of enhancing low resolution photos to high resolution photographs with minimal loss of image quality appears to be known as super resolution. The most effective method of splendid decision is primarily relies on interpolation, that anticipates the high resolution pixels depending on their surroundings. This technique has a reduced computational cost and a great real-time performance, Bilinear or bicubic interpolation methods comes under simple interpolation methods, for example, to produce smooth images with ringing and jagged artefacts, which leads to poor iterative reconstruction quality. When there are few or no accessible input images, and whenever the needed magnification power is high, the quality of the reconstructed image quickly declines. The result in these circumstances could be overly straightforward and omit crucial high frequency data. In this article, we describe how a learning-based super resolution reconstruction method can utilize the prior knowledge in low resolution images and the prior information collected through learning. Even with a high magnification factor, excellent reconstruction results can be produced. Learning-based super resolution has thus emerged as among the most efficient methods for a variety of real-world applications.
