Spatially-Variant Degradation Model for Dataset-free Super-resolution

TL;DR

This paper introduces a novel spatially-variant degradation model for dataset-free blind image super-resolution, using a small parameter set and probabilistic inference to outperform existing methods.

Contribution

It is the first to explicitly model pixel-wise spatially-variant degradation in dataset-free BISR, reducing parameters and improving accuracy.

Findings

Achieves an average of 1 dB improvement over state-of-the-art methods.

Uses a small number of learnable parameters for degradation kernels.

Employs Monte Carlo EM for kernel inference.

Abstract

This paper focuses on the dataset-free Blind Image Super-Resolution (BISR). Unlike existing dataset-free BISR methods that focus on obtaining a degradation kernel for the entire image, we are the first to explicitly design a spatially-variant degradation model for each pixel. Our method also benefits from having a significantly smaller number of learnable parameters compared to data-driven spatially-variant BISR methods. Concretely, each pixel's degradation kernel is expressed as a linear combination of a learnable dictionary composed of a small number of spatially-variant atom kernels. The coefficient matrices of the atom degradation kernels are derived using membership functions of fuzzy set theory. We construct a novel Probabilistic BISR model with tailored likelihood function and prior terms. Subsequently, we employ the Monte Carlo EM algorithm to infer the degradation kernels for each pixel. Our method achieves a significant improvement over other state-of-the-art BISR methods, with an average improvement of 1 dB (2x).Code will be released at https://github.com/shaojieguoECNU/SVDSR.