講演情報
[8p-A23-7]A Split-Step Propagation-Based Simulation Method for Training Deep Turbulence Removal Models
〇(M2)Guan-Lin Huang1, Jia-Han Li1 (1.ESOE NTU)
キーワード：
Split-step propagation、Turbulence removal、Image restoration
In long-range imaging in atmospheric environment or short-range underwater observation systems, light propagating through non-uniform media is often degraded by turbulence-induced disturbances. These disturbances mainly arise from refractive-index fluctuations caused by temperature gradients, density variations, or flow-field motion, leading to phase perturbations, beam wander, intensity scintillation, and image spreading . Unlike general image degradation, turbulence involves geometric displacement, spatially varying blur, and sensor noise, making modeling and restoration more challenging . Therefore, this study uses the split-step propagation method to generate physically consistent turbulence-degraded images for training a deep turbulence removal model, aiming to improve image clarity, structural integrity, and recognizability.