DEFormer: DCT-driven Enhancement Transformer for Low-light Image and Dark Vision

2023-09-13

arXiv.org

AI Summary

To address the challenges of recovering fine details and accurate color in dark regions of low-light images, this paper proposes a cross-domain enhancement Transformer framework based on discrete cosine transform (DCT) domain modeling. The method introduces three key innovations: (1) a learnable DCT transform module for adaptive frequency-domain representation; (2) a curvature-aware frequency enhancement (CFE) mechanism to selectively amplify responses in critical frequency bands; and (3) an RGB–frequency domain cross-fusion module (CDF) enabling synergistic spatial–frequency feature modeling. Evaluated on the LOL and MIT-Adobe FiveK datasets, the approach achieves state-of-the-art performance in PSNR and SSIM metrics, and significantly outperforms existing methods on downstream low-light object detection tasks. These results empirically validate the effectiveness of frequency-domain priors in guiding low-light image enhancement.

Abstract

Low-light image enhancement restores the colors and details of a single image and improves high-level visual tasks. However, restoring the lost details in the dark area is still a challenge relying only on the RGB domain. In this paper, we delve into frequency as a new clue into the model and propose a DCT-driven enhancement transformer (DEFormer) framework. First, we propose a learnable frequency branch (LFB) for frequency enhancement contains DCT processing and curvature-based frequency enhancement (CFE) to represent frequency features. Additionally, we propose a cross domain fusion (CDF) to reduce the differences between the RGB domain and the frequency domain. Our DEFormer has achieved superior results on the LOL and MIT-Adobe FiveK datasets, improving the dark detection performance.

Problem

• Low-light image enhancement
• Traditional enhancement techniques
• Detail and color recovery

Innovation

• DEFormer
• Discrete Cosine Transform
• Low-light Image Enhancement