Spatial-Guided Diffusion with State Space Modeling for Low-Light Image Enhancement

Low-light image enhancement remains a challenging task due to non-uniform illumination, information loss, and noise amplification. In this paper, we propose a novel spatialguided multi-step diffusion framework that adaptively modulates noise distribution based on the importance of image regions. Specifically, a spatial-guided module is first employed to estimate the significance of different areas in the input image. This map then guides a progressive multi-step noise addition process that simulates region-aware degradation. The resulting noisy images are sequentially denoised using a lightweight SSM-based model that preserves global consistency and texture fidelity. Finally, a second-stage SSD model further refines the image to produce high-quality results. Extensive experiments on multiple benchmark datasets demonstrate that our method achieves competitive performance against state-of-the-art approaches, with superior results in terms of PSNR, SSIM, and LPIPS. The proposed architecture offers a promising solution for real-world low-light enhancement applications in photography, surveillance, and autonomous vision.