The paper introduces RetinexMamba, an architecture for low-light image enhancement that combines traditional Retinex methods with deep learning techniques. Traditional Retinex methods, which decompose images into illumination and reflection components, struggle with noise management and detail preservation in low-light conditions. Deep learning models like Retinexformer enhance illumination estimation through self-attention mechanisms but face challenges with interpretability and suboptimal enhancement effects. RetinexMamba integrates the physical intuitiveness of traditional Retinex methods with the computational efficiency of State Space Models (SSMs). It features innovative illumination estimators and damage restorers that maintain image quality during enhancement. The architecture replaces the IG-MSA in Retinexformer with a Fused-Attention mechanism, improving interpretability. Experimental results on the LOL dataset show that RetinexMamba outperforms existing deep learning approaches based on Retinex theory in both quantitative and qualitative metrics, demonstrating its effectiveness and superiority in low-light image enhancement.The paper introduces RetinexMamba, an architecture for low-light image enhancement that combines traditional Retinex methods with deep learning techniques. Traditional Retinex methods, which decompose images into illumination and reflection components, struggle with noise management and detail preservation in low-light conditions. Deep learning models like Retinexformer enhance illumination estimation through self-attention mechanisms but face challenges with interpretability and suboptimal enhancement effects. RetinexMamba integrates the physical intuitiveness of traditional Retinex methods with the computational efficiency of State Space Models (SSMs). It features innovative illumination estimators and damage restorers that maintain image quality during enhancement. The architecture replaces the IG-MSA in Retinexformer with a Fused-Attention mechanism, improving interpretability. Experimental results on the LOL dataset show that RetinexMamba outperforms existing deep learning approaches based on Retinex theory in both quantitative and qualitative metrics, demonstrating its effectiveness and superiority in low-light image enhancement.