Accelerating Atmospheric Turbulence Simulation via Learned Phase-to-Space Transform

TL;DR

This paper introduces a learned phase-to-space transform that accelerates atmospheric turbulence simulation by 300 to 1000 times while maintaining key turbulence statistics, enabling faster development of turbulence mitigation algorithms.

Contribution

The paper proposes a novel phase-to-space transform that reformulates turbulence simulation, leveraging learned basis functions and a lightweight network for rapid phase-to-space conversion.

Findings

Achieves 300x to 1000x speedup over traditional simulators.

Preserves essential turbulence statistics in simulations.

Enables faster development of turbulence mitigation algorithms.

Abstract

Fast and accurate simulation of imaging through atmospheric turbulence is essential for developing turbulence mitigation algorithms. Recognizing the limitations of previous approaches, we introduce a new concept known as the phase-to-space (P2S) transform to significantly speed up the simulation. P2S is build upon three ideas: (1) reformulating the spatially varying convolution as a set of invariant convolutions with basis functions, (2) learning the basis function via the known turbulence statistics models, (3) implementing the P2S transform via a light-weight network that directly convert the phase representation to spatial representation. The new simulator offers 300x -- 1000x speed up compared to the mainstream split-step simulators while preserving the essential turbulence statistics.