The article presents an efficient network for Low-Light Image Enhancement (LLIE) named Horizontal/Vertical-Intensity (HVI) and Color and Intensity Decoupling Network (CIDNet). Traditional methods use Deep Neural Networks (DNNs) in sRGB and HSV color spaces, but this can lead to instability and color artifacts in the enhanced images. The proposed HVI color space decouples brightness and color to mitigate this instability, and the CIDNet processes the decoupled brightness and color, suppressing noise in low-light images. The study shows that the proposed method outperforms existing ones in various experiments.
Publication date: 9 Feb 2024
Project Page: https://github.com/Fediory/HVI-CIDNet
Paper: https://arxiv.org/pdf/2402.05809
