Learning the Degradation Distribution for Blind Image Super-Resolution

TL;DR

This paper introduces a probabilistic degradation model for blind image super-resolution that captures stochastic degradations, leading to more diverse training pairs and improved SR performance across datasets.

Contribution

It proposes a novel probabilistic degradation model (PDM) that learns the distribution of degradations as a stochastic process, enhancing the realism and diversity of training data for SR.

Findings

PDM models more diverse degradations than deterministic models.

SR models trained with PDM outperform those trained with traditional methods.

The approach improves generalization across different datasets.

Abstract

Synthetic high-resolution (HR) \& low-resolution (LR) pairs are widely used in existing super-resolution (SR) methods. To avoid the domain gap between synthetic and test images, most previous methods try to adaptively learn the synthesizing (degrading) process via a deterministic model. However, some degradations in real scenarios are stochastic and cannot be determined by the content of the image. These deterministic models may fail to model the random factors and content-independent parts of degradations, which will limit the performance of the following SR models. In this paper, we propose a probabilistic degradation model (PDM), which studies the degradation $\mathbf{D}$ as a random variable, and learns its distribution by modeling the mapping from a priori random variable $\mathbf{z}$ to $\mathbf{D}$. Compared with previous deterministic degradation models, PDM could model more diverse degradations and generate HR-LR pairs that may better cover the various degradations of test images, and thus prevent the SR model from over-fitting to specific ones. Extensive experiments have demonstrated that our degradation model can help the SR model achieve better performance on different datasets. The source codes are released at \url{git@github.com:greatlog/UnpairedSR.git}.