You Only Need One Color Space: An Efficient Network for Low-light Image Enhancement

8 Feb 2024  ·  Qingsen Yan, Yixu Feng, Cheng Zhang, Pei Wang, Peng Wu, Wei Dong, Jinqiu Sun, Yanning Zhang ·

Low-Light Image Enhancement (LLIE) task tends to restore the details and visual information from corrupted low-light images. Most existing methods learn the mapping function between low/normal-light images by Deep Neural Networks (DNNs) on sRGB and HSV color space. Nevertheless, enhancement involves amplifying image signals, and applying these color spaces to low-light images with a low signal-to-noise ratio can introduce sensitivity and instability into the enhancement process. Consequently, this results in the presence of color artifacts and brightness artifacts in the enhanced images. To alleviate this problem, we propose a novel trainable color space, named Horizontal/Vertical-Intensity (HVI). It not only decouples brightness and color from RGB channels to mitigate the instability during enhancement but also adapts to low-light images in different illumination ranges due to the trainable parameters. Further, we design a novel Color and Intensity Decoupling Network (CIDNet) with two branches dedicated to processing the decoupled image brightness and color in the HVI space. Within CIDNet, we introduce the Lightweight Cross-Attention (LCA) module to facilitate interaction between image structure and content information in both branches, while also suppressing noise in low-light images. Finally, we conducted 22 quantitative and qualitative experiments to show that the proposed CIDNet outperforms the state-of-the-art methods on 11 datasets. The code is available at https://github.com/Fediory/HVI-CIDNet.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Low-Light Image Enhancement DICM CIDNet NIQE 3.79 # 1
BRISQUE 21.47 # 1
Low-Light Image Enhancement LIME CIDNet NIQE 4.13 # 1
BRISQUE 16.25 # 1
Low-Light Image Enhancement LOL CIDNet-Normal Average PSNR 23.500 # 18
SSIM 0.870 # 6
LPIPS 0.086 # 5
FLOPS (G) 7.57 # 2
Params (M) 1.88 # 4
SSIM (sRGB) 0.870 # 2
Low-Light Image Enhancement LOL CIDNet Average PSNR 28.141 # 1
SSIM 0.889 # 3
LPIPS 0.079 # 2
FLOPS (G) 7.57 # 2
Params (M) 1.88 # 4
SSIM (sRGB) 0.889 # 1
Low-light Image Deblurring and Enhancement LOL-Blur CIDNet SSIM 0.890 # 1
LPIPS 0.120 # 1
Average PSNR 26.572 # 1
Low-Light Image Enhancement LOLv2 CIDNet Average PSNR 28.134 # 4
SSIM 0.892 # 4
LPIPS 0.101 # 4
Low-Light Image Enhancement LOL-v2 CIDNet Average PSNR 24.111 # 1
SSIM 0.868 # 2
LPIPS 0.108 # 1
Low-Light Image Enhancement LOL-v2-synthetic CIDNet PSNR 25.705 # 2
SSIM 0.942 # 2
LPIPS 0.045 # 1
Low-Light Image Enhancement LOLv2-synthetic CIDNet Average PSNR 29.566 # 1
SSIM 0.950 # 1
LPIPS 0.040 # 2
Low-Light Image Enhancement MEF CIDNet NIQE 3.56 # 2
BRISQUE 13.77 # 1
Low-Light Image Enhancement NPE CIDNet NIQE 3.74 # 1
BRISQUE 18.92 # 1
Image Enhancement SICE-Grad CIDNet Average PSNR 13.446 # 1
SSIM 0.648 # 1
LPIPS 0.318 # 1
Image Enhancement SICE-Mix CIDNet Average PSNR 13.425 # 1
SSIM 0.636 # 1
LPIPS 0.362 # 1
Low-Light Image Enhancement Sony-Total-Dark CIDNet Average PSNR 22.904 # 1
SSIM 0.676 # 1
LPIPS 0.411 # 1
Low-Light Image Enhancement VV CIDNet NIQE 3.21 # 1
BRISQUE 30.63 # 1

Methods