Zero-Shot Low-Light-Level Image Enhancement Algorithm Based on Atmospheric Scattering Model

Low-light image enhancement remains a pivotal area of focus within the realm of image processing. While current learning-based enhancement techniques have shown advancements in managing noise or color distortion, they often fail to strike an optimal balance between these two aspects and are constrained by the availability and quality of training data. In this research, we introduce a novel zero-shot learning method for low-light image enhancement, named ZRASM, which leverages the Atmospheric Scattering Model (ASM) and integrates a denoising network grounded in the Noise2Noise theory. ZRASM is capable of noise reduction and color restoration without the need for external training data, operating solely from a single input image. Our experimental results indicate that ZRASM surpasses several state-of-the-art models in terms of image enhancement performance.