Meta-Learned Kernel For Blind Super-Resolution Kernel Estimation

TL;DR

This paper introduces MetaKernelGAN, a meta-learned kernel generator that significantly speeds up and improves the accuracy of blind super-resolution by better estimating degradation kernels through a learned prior.

Contribution

It proposes a meta-learning approach for kernel estimation in blind super-resolution, enabling faster adaptation and higher fidelity results compared to existing methods.

Findings

Achieves state-of-the-art kernel estimation accuracy.

Provides 14.24 to 102.1x faster inference speed.

Improves super-resolution image quality with better kernel estimates.

Abstract

Recent image degradation estimation methods have enabled single-image super-resolution (SR) approaches to better upsample real-world images. Among these methods, explicit kernel estimation approaches have demonstrated unprecedented performance at handling unknown degradations. Nonetheless, a number of limitations constrain their efficacy when used by downstream SR models. Specifically, this family of methods yields i) excessive inference time due to long per-image adaptation times and ii) inferior image fidelity due to kernel mismatch. In this work, we introduce a learning-to-learn approach that meta-learns from the information contained in a distribution of images, thereby enabling significantly faster adaptation to new images with substantially improved performance in both kernel estimation and image fidelity. Specifically, we meta-train a kernel-generating GAN, named MetaKernelGAN, on a range of tasks, such that when a new image is presented, the generator starts from an informed kernel estimate and the discriminator starts with a strong capability to distinguish between patch distributions. Compared with state-of-the-art methods, our experiments show that MetaKernelGAN better estimates the magnitude and covariance of the kernel, leading to state-of-the-art blind SR results within a similar computational regime when combined with a non-blind SR model. Through supervised learning of an unsupervised learner, our method maintains the generalizability of the unsupervised learner, improves the optimization stability of kernel estimation, and hence image adaptation, and leads to a faster inference with a speedup between 14.24 to 102.1x over existing methods.