This paper proposes a novel low-light image enhancement method called CIDNet, which uses a new color space called Horizontal/Vertical-Intensity (HVI) to decouple brightness and color information. The HVI color space is designed to adapt to different illumination conditions and includes trainable parameters to improve the enhancement process. CIDNet consists of two branches: an HV-branch for color information and an intensity-branch for brightness. The method also incorporates a Lighten Cross-Attention (LCA) module to facilitate interaction between the branches and suppress noise. The proposed method outperforms existing state-of-the-art methods on 11 datasets in terms of PSNR, SSIM, and LPIPS metrics. The HVI color space and CIDNet architecture are validated through extensive experiments, demonstrating their effectiveness in enhancing low-light images while preserving natural colors. The method achieves a good balance between effectiveness and efficiency, with a small parameter count (1.88M) and computational load (7.57G), making it suitable for deployment on edge devices. The paper also includes ablation studies and comparisons with other methods, showing that the HVI color space and LCA module significantly improve the performance of low-light image enhancement.This paper proposes a novel low-light image enhancement method called CIDNet, which uses a new color space called Horizontal/Vertical-Intensity (HVI) to decouple brightness and color information. The HVI color space is designed to adapt to different illumination conditions and includes trainable parameters to improve the enhancement process. CIDNet consists of two branches: an HV-branch for color information and an intensity-branch for brightness. The method also incorporates a Lighten Cross-Attention (LCA) module to facilitate interaction between the branches and suppress noise. The proposed method outperforms existing state-of-the-art methods on 11 datasets in terms of PSNR, SSIM, and LPIPS metrics. The HVI color space and CIDNet architecture are validated through extensive experiments, demonstrating their effectiveness in enhancing low-light images while preserving natural colors. The method achieves a good balance between effectiveness and efficiency, with a small parameter count (1.88M) and computational load (7.57G), making it suitable for deployment on edge devices. The paper also includes ablation studies and comparisons with other methods, showing that the HVI color space and LCA module significantly improve the performance of low-light image enhancement.