RSEND: Retinex-based Squeeze and Excitation Network with Dark Region Detection for Efficient Low Light Image Enhancement

TL;DR

Low-light image enhancement demands improved visibility without heavy computation. The authors introduce RSEND, a one-stage Retinex-based network that decomposes input $S$ into $S = R \circ I$, applies a dark-region attention path, and uses a compact, SE-guided illumination enhancement with a denoiser, achieving end-to-end training. RSEND yields PSNR gains up to $4.03$ dB and SSIM up to $0.912$ across datasets with only 0.41M parameters and substantially lower FLOPs, outperforming CNN-based SOTA and even some transformer-based models on LOL-v2-real. The approach demonstrates that careful Retinex decomposition, region-aware illumination, and channel-wise feature recalibration enable high-quality, efficient low-light enhancement suitable for mobile and edge deployment.

Abstract

Images captured under low-light scenarios often suffer from low quality. Previous CNN-based deep learning methods often involve using Retinex theory. Nevertheless, most of them cannot perform well in more complicated datasets like LOL-v2 while consuming too much computational resources. Besides, some of these methods require sophisticated training at different stages, making the procedure even more time-consuming and tedious. In this paper, we propose a more accurate, concise, and one-stage Retinex theory based framework, RSEND. RSEND first divides the low-light image into the illumination map and reflectance map, then captures the important details in the illumination map and performs light enhancement. After this step, it refines the enhanced gray-scale image and does element-wise matrix multiplication with the reflectance map. By denoising the output it has from the previous step, it obtains the final result. In all the steps, RSEND utilizes Squeeze and Excitation network to better capture the details. Comprehensive quantitative and qualitative experiments show that our Efficient Retinex model significantly outperforms other CNN-based models, achieving a PSNR improvement ranging from 0.44 dB to 4.2 dB in different datasets and even outperforms transformer-based models in the LOL-v2-real dataset.

Figures (6)

• Figure 1: Comparison of our RSEND against previous CNN based state-of-the-art methods including SIDchen2019seeing, RetinexNet (RTXNet)wei2018deep, EnGANjiang2021enlightengan, DRBNyang2021band, KinDzhang2019kindling on three datasets: LOL-v1, LOL-v2-real, and LOL-v2-syn. Our RSEND flow achieves 1.69 dB to 3.63 dB improvements over the best previous works in terms of PSNR, as indicated by the orange +dB annotations.
• Figure 2: The proposed framework of RSEND. Our network consists of five subnets: a Decom-Net, a Dark Region Detection-Net, an Enhancer-Net, a Refinement-Net, and a Denoiser. The Decom-Net decomposes the low-light image into a reflectance map and an illumination map based on the Retinex theory. The Dark Region Detection-Net means to find the regions that need to be enhanced more. The Enhancer-Net functions to illuminate the illumination map. The Refinement-Net aims to adjust contrasts and fine-tune the details. In the end, Denoiser performs denoising to get clean and visually pleasing output.
• Figure 3: Effect of Dark Region Detection. The left image shows the illumination map before dark region detection, while the right image demonstrates enhanced focus on darker areas after applying the module.
• Figure 4: Visual comparisons with Retinexformer cai2023retinexformer, DC-Net zhang2022deep, EnGAN jiang2021enlightengan, MIR-Net zamir2020learning, RetinexNet wei2018deep, Zero-DCE guo2020zero our RSEND performs better.
• Figure 5: Ablation Study of the effect of each model component