Machine-learning-based image quality prediction from optical turbulence and meteorological observations

Atmospheric turbulence can significantly impact the quality of an image by causing distortions during image acquisition. The turbulence can affect the image sharpness and contrast. We will quantify atmospheric impacts on image quality by using standard image quality metrics for measuring image blur (such as the variance of the Laplacian). In this paper we will utilize meteorological and contrast targetboard observations to train machinelearning models. We utilize approximately two years of optical turbulence data and meteorological data along a coastal path to train and test the models. A comparison between various ML algorithms, such as XGBoost (Extreme Gradient Boosting) and LightGBM (Light Gradient Boosting Machine), is performed to determine which best predicts the image quality metrics.