DWA: Differential Wavelet Amplifier for Image Super-Resolution

TL;DR

This paper presents Differential Wavelet Amplifier (DWA), a novel module that enhances wavelet-based image super-resolution models by refining feature extraction and reducing computational costs, leading to improved SR performance.

Contribution

The introduction of DWA as a drop-in module improves wavelet-based SR models by leveraging filter differences to enhance feature extraction and eliminate the need for traditional DWT.

Findings

DWA improves SR model performance on classical tasks.

Integration of DWA reduces model size and computation.

DWA enables direct application to image space, bypassing DWT.

Abstract

This work introduces Differential Wavelet Amplifier (DWA), a drop-in module for wavelet-based image Super-Resolution (SR). DWA invigorates an approach recently receiving less attention, namely Discrete Wavelet Transformation (DWT). DWT enables an efficient image representation for SR and reduces the spatial area of its input by a factor of 4, the overall model size, and computation cost, framing it as an attractive approach for sustainable ML. Our proposed DWA model improves wavelet-based SR models by leveraging the difference between two convolutional filters to refine relevant feature extraction in the wavelet domain, emphasizing local contrasts and suppressing common noise in the input signals. We show its effectiveness by integrating it into existing SR models, e.g., DWSR and MWCNN, and demonstrate a clear improvement in classical SR tasks. Moreover, DWA enables a direct application of DWSR and MWCNN to input image space, reducing the DWT representation channel-wise since it omits traditional DWT.