The field of low-light image processing is rapidly advancing with a focus on developing innovative methods to improve image quality and enhance visual understanding. Recent research has emphasized the importance of bridging the gap between low-light enhancement and high-level visual understanding, leading to the development of generalized frameworks that can effectively address diverse causes of low-light degradation. Another notable trend is the incorporation of advanced techniques such as generative latent kernel modeling, Schrödinger bridge theory, and wavelet transforms to tackle complex image restoration tasks like blind motion deblurring and image dehazing. Furthermore, there is a growing interest in exploring the potential of non-traditional imaging modalities like event cameras for radiance field reconstruction in low-light environments. Noteworthy papers in this area include:

• A method proposing a generalized bridge between low-light enhancement and low-light understanding, leveraging pretrained generative diffusion models for zero-shot generalization performance.
• A framework utilizing a deep generative model to encode the kernel prior for blind motion deblurring, enhancing the initialization and robustness of the blur kernel estimation.
• A novel unpaired dehazing approach based on the Schrödinger Bridge, enabling optimal transport mappings from hazy to clear images for high-quality results.