Striving for Faster and Better: A One-Layer Architecture with Auto Re-parameterization for Low-Light Image Enhancement

TL;DR

This paper introduces a novel one-layer neural network architecture with auto re-parameterization for low-light image enhancement, achieving high visual quality and superior computational efficiency across multiple hardware platforms.

Contribution

It proposes a minimalist single-layer model with re-parameterization and hierarchical search, significantly improving both performance and speed in low-light image enhancement.

Findings

Outperforms recent methods in visual quality and efficiency

Achieves faster processing on CPU, GPU, NPU, DSP platforms

Maintains excellent enhancement quality with a simple model structure

Abstract

Deep learning-based low-light image enhancers have made significant progress in recent years, with a trend towards achieving satisfactory visual quality while gradually reducing the number of parameters and improving computational efficiency. In this work, we aim to delving into the limits of image enhancers both from visual quality and computational efficiency, while striving for both better performance and faster processing. To be concrete, by rethinking the task demands, we build an explicit connection, i.e., visual quality and computational efficiency are corresponding to model learning and structure design, respectively. Around this connection, we enlarge parameter space by introducing the re-parameterization for ample model learning of a pre-defined minimalist network (e.g., just one layer), to avoid falling into a local solution. To strengthen the structural representation, we define a hierarchical search scheme for discovering a task-oriented re-parameterized structure, which also provides powerful support for efficiency. Ultimately, this achieves efficient low-light image enhancement using only a single convolutional layer, while maintaining excellent visual quality. Experimental results show our sensible superiority both in quality and efficiency against recently-proposed methods. Especially, our running time on various platforms (e.g., CPU, GPU, NPU, DSP) consistently moves beyond the existing fastest scheme. The source code will be released at https://github.com/vis-opt-group/AR-LLIE.