Imagify: Reconstruction of High- Resolution Images from Degraded Images

Abstract

Abstract: Image Denoising simply uses photos with noise and the U-Net architecture. In this study, we demonstrate that the unet design, which is based on convolutional and deconvolutional neural networks or transpose convolutional neural networks, is quite effective at reducing image noise. The assignment falls under abroad category of problems on distribution of posterior probabilities P(theta | X) represents the probability of theparameter theta given the evidence X. These techniques yield great results, but their practical application might be problematic due to issues like the expensive forward and adjoint operators' computations and the challenging hyper parameter selection.