Researchers have introduced TCA-Net, a novel network designed for low-light image enhancement that focuses on reliable fusion of intensity and chromaticity information. Unlike previous methods that use fixed quotas for attention, TCA-Net employs a confidence threshold that adaptively retains high-confidence cross-stream interactions. The network also incorporates a Phase-guided Fourier Interaction Module for brightness initialization and a Decoupled Dual-Stream Guidance Module to prevent chromaticity leakage, along with Scale-Aware Consistency Regularization for structural robustness. Experiments show TCA-Net achieves competitive accuracy and improved color fidelity with a compact parameter size. AI
IMPACT Introduces a novel approach to low-light image enhancement, potentially improving performance and efficiency in applications requiring high-quality visual fidelity.
RANK_REASON The cluster contains a research paper detailing a new method and network for image enhancement.
- Decoupled Dual-Stream Guidance Module
- LOL-v1
- LOL-v2
- LSRW-Huawei
- Phase-guided Fourier Interaction Module
- Scale-Aware Consistency Regularization
- Sony-Total-Dark
- TCA-Net
- Thresholded Cross-Attention
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
