Real-Time Dense Field Phase-to-Space Simulation of Imaging Through Atmospheric Turbulence

Numerical simulation of atmospheric turbulence is one the biggest bottlenecks in developing computational techniques for solving inverse problem long-range imaging. The classical split-step method based upon numerical wave propagation which splits path into many segments and propagates every pixel each segment individually via Fresnel integral. This repeated evaluation becomes increasingly time-consuming larger images. As a result, often done only on sparse grid points followed by an interpolation to other pixels. Even so, computation expensive real-time applications. In this paper, we present new that enables real-time processing over xmlns:xlink="http://www.w3.org/1999/xlink">dense points. Building recently developed multi-aperture model phase-to-space transform, overcome memory bottleneck drawing random samples from Zernike correlation tensor. We show cross-correlation modes has insignificant contribution statistics samples. By approximating these blocks tensor, restore homogeneity tensor then Fourier-based sampling. On $512\times 512$ image, simulator achieves 0.025 seconds per frame dense field. notation="LaTeX">$3840 \times 2160$ image would have taken 13 hours simulate using method, can run at approximately 60 frame.