Thresholded Cross-Attention for Reliable Intensity-Chromaticity Fusion in Low-Light Image Enhancement

2026-07-15

AI Summary

This work addresses the challenge in low-light image enhancement where unreliable fusion of intensity and chrominance information often compromises the balance among noise suppression, color fidelity, and computational efficiency. To this end, the authors propose TCA-Net, which introduces a thresholded cross-attention mechanism in the HVI color space to enable input- and layer-adaptive, high-confidence cross-stream feature fusion, replacing conventional fixed Top-K sparsification strategies. Additionally, a phase-guided Fourier interaction module and a decoupled dual-stream guidance module are designed, leveraging frequency-domain initialization and residual guidance to effectively suppress chrominance leakage and enhance structural consistency. The method achieves state-of-the-art performance in restoration accuracy and color fidelity while maintaining a compact model size on benchmark datasets including LOL-v1/v2, Sony-Total-Dark, and LSRW-Huawei.

Abstract

Low-Light Image Enhancement (LLIE) requires a careful balance among noise suppression, color fidelity, and efficiency. Recent HVI-based methods alleviate color entanglement by decoupling intensity and chromaticity, yet how reliably the two streams are fused again is an overlooked factor that largely determines the final quality. We observe that the confidence of cross-stream attention is strongly layer-dependent, so the fixed-quota selection of Top-K sparse attention is mismatched to it, discarding informative dependencies in some layers while retaining noisy ones in others. Motivated by this observation, we propose TCA-Net, a network built around Thresholded Cross-Attention that targets reliable intensity-chromaticity fusion in the HVI space rather than introducing yet another color representation. At its core, TCA replaces the rigid Top-K quota with a fixed confidence threshold whose retained cardinality is input- and layer-adaptive, retaining only high-confidence cross-stream interactions while suppressing unreliable ones. Around this core, two complementary designs clean up the fusion before and after it: a Phase-guided Fourier Interaction Module provides a structure-aware brightness initialization for the intensity stream prior to fusion, and a Decoupled Dual-Stream Guidance Module constructs residual intensity features to suppress chromaticity leakage during reconstruction. A Scale-Aware Consistency Regularization further improves structural robustness under scale perturbations during training. Extensive experiments on LOL-v1, LOL-v2, Sony-Total-Dark, and LSRW-Huawei demonstrate that TCA-Net delivers competitive restoration accuracy, improved color fidelity, and a compact parameter size.

Problem

• Low-Light Image Enhancement
• Intensity-Chromaticity Fusion
• Cross-Attention
• Color Fidelity
• Noise Suppression

Innovation

• Thresholded Cross-Attention
• Intensity-Chromaticity Fusion
• HVI Space
• Adaptive Sparsity
• Color Fidelity