LTCF-Net: A Transformer-Enhanced Dual-Channel Fourier Framework for Low-Light Image Restoration

TL;DR

LTCF-Net is introduced, a novel network architecture designed for enhancing low-light images that outperforms current state-of-the-art approaches across multiple evaluation metrics and datasets, achieving more natural color restoration and a balanced brightness distribution.

Abstract

We introduce LTCF-Net, a novel network architecture designed for enhancing low-light images. Unlike Retinex-based methods, our approach utilizes two color spaces - LAB and YUV - to efficiently separate and process color information, by leveraging the separation of luminance from chromatic components in color images. In addition, our model incorporates the Transformer architecture to comprehensively understand image content while maintaining computational efficiency. To dynamically balance the brightness in output images, we also introduce a Fourier transform module that adjusts the luminance channel in the frequency domain. This mechanism could uniformly balance brightness across different regions while eliminating background noises, and thereby enhancing visual quality. By combining these innovative components, LTCF-Net effectively improves low-light image quality while keeping the model lightweight. Experimental results demonstrate that our method outperforms current state-of-the-art approaches across multiple evaluation metrics and datasets, achieving more natural color restoration and a balanced brightness distribution.

Figures (8)

• Figure 1: Model Pipeline. Our main models are Multi-header Self-attention (MHSA) Block, Channel Denoising (CD) Block, Multi-stage Squeeze and Excited Fusion (MSEF) Block and Fourier Branch Processing (FBP) Block. The individual submodules can be seen in Fig.2.
• Figure 2: Submodules in our model. (a) Multi-headed Self-attention (MHSA) Block use multi-head attention mechanism to acquire features. (b) Multi-stage Squeeze and Excited Fusion (MSEF) Block, Multi-stage processing and capture of global and local features for image recovery. (c) Channel Denoiser (CD) Block, step and jump connections based on U-shaped networks are used for denoising. (d) Fourier Bright Processing (FBP) Block uses Fourier transform to de-noise the light information.
• Figure 3: Results on LOL-v113(top) and LOL-v2-real12(bottom). Our method effectively enhances the visibility and preserves the color.
• Figure 4: Results on SID15(top) and SMID48(bottom). Previous methods either output incorrect colors or have strong noise. Our method effectively improves the visibility of the image and preserves the rich details.
• Figure 5: Results on SDSD-indoor 14(top) and SDSD-outdoor 14(bottom). Benchmark methods either output incorrect colors or have strong noise, while the proposed method effectively improves the visibility of the image and preserves the rich details.