A signal processing interpretation of noise-reduction convolutional neural networks

TL;DR

This paper bridges signal processing and deep learning by providing a unified theoretical framework for encoding-decoding CNNs used in noise reduction, enhancing understanding and guiding the design of more effective architectures.

Contribution

It introduces a signal processing interpretation of noise-reduction CNNs based on deep convolutional framelets, unifying diverse architectures under a common theoretical framework.

Findings

Provides a signal processing perspective on CNNs for noise reduction

Unifies various CNN architectures using deep convolutional framelets

Offers guidance for designing robust and efficient CNNs

Abstract

Encoding-decoding CNNs play a central role in data-driven noise reduction and can be found within numerous deep-learning algorithms. However, the development of these CNN architectures is often done in ad-hoc fashion and theoretical underpinnings for important design choices is generally lacking. Up to this moment there are different existing relevant works that strive to explain the internal operation of these CNNs. Still, these ideas are either scattered and/or may require significant expertise to be accessible for a bigger audience. In order to open up this exciting field, this article builds intuition on the theory of deep convolutional framelets and explains diverse ED CNN architectures in a unified theoretical framework. By connecting basic principles from signal processing to the field of deep learning, this self-contained material offers significant guidance for designing robust and efficient novel CNN architectures.