Discovering Distinctive "Semantics" in Super-Resolution Networks

TL;DR

This paper uncovers that super-resolution networks learn distinctive semantic features related to image degradation rather than content, revealing new insights into their working mechanisms and potential applications.

Contribution

It is the first to identify and analyze the semantic representations related to degradation in SR networks, providing a new understanding of their internal mechanisms.

Findings

Deep SR networks encode degradation-related semantics.

Adversarial learning and global residuals affect semantic extraction.

DDR can be used for distortion identification and evaluating generalization.

Abstract

Image super-resolution (SR) is a representative low-level vision problem. Although deep SR networks have achieved extraordinary success, we are still unaware of their working mechanisms. Specifically, whether SR networks can learn semantic information, or just perform complex mapping function? What hinders SR networks from generalizing to real-world data? These questions not only raise our curiosity, but also influence SR network development. In this paper, we make the primary attempt to answer the above fundamental questions. After comprehensively analyzing the feature representations (via dimensionality reduction and visualization), we successfully discover the distinctive "semantics" in SR networks, i.e., deep degradation representations (DDR), which relate to image degradation instead of image content. We show that a well-trained deep SR network is naturally a good descriptor of degradation information. Our experiments also reveal two key factors (adversarial learning and global residual) that influence the extraction of such semantics. We further apply DDR in several interesting applications (such as distortion identification, blind SR and generalization evaluation) and achieve promising results, demonstrating the correctness and effectiveness of our findings.