Multi-Step Adaptive Deconvolution via Multi-View for Image Enhancement in Scattering Media

Imaging through scattering media under natural illumination is a critical challenge in applications such as autonomous driving, underwater detection, and aviation safety. We propose a novel multi-view multi-step deconvolution image enhancement method for imaging in scattering media, which combines image background noise suppression, detail enhancement, and super-resolution. It integrates the advantages of dark channel prior preprocessing and adaptive deconvolution. A camera array-based common field of view multi-frame imaging model is established to address the challenges of long acquisition time and difficulty in registering moving objects in single-camera systems, thereby enabling multi-frame image acquisition of objects. Then, a multi-step adaptive deconvolution scattering imaging algorithm is designed for image enhancement, background noise suppression, and super-resolution imaging of objects, enabling multi-frame image fusion. The proposed approach tackles the issue of low signal-to-noise ratio in scattering environments, achieving high-quality passive imaging in complex scattering scenes. Finally, the proposed method is validated through simulations, laboratory-equivalent experiments, and real-world experiments, demonstrating its ability to achieve high-quality imaging in complex scattering scenarios with natural illumination, especially in heavy haze environments with limited camera quantities. The results show that when comparing with the existing methods under artificial heavy, moderate, and light haze conditions, the proposed method improves PSNR by 2.24%, 1.83%, and 2.95%, and SSIM by 10.93%, 5.21%, and 5.53%, respectively.