Imaging through Atmospheric Turbulence

Overview

Imaging through atmospheric turbulence is a fundamental problem for long-range imaging systems. Purdue i2Lab has specialities in several aspects of the subject:

Wave optics theory
Turbulence simulators
Forward imaging models that can be used for deep neural networks
Image reconstruction algorithms
Object recognition

Tutorial

ICIP 2022 and CVPR 2022
Instructors: Stanley Chan and Nick Chimitt

Recordings

Purdue Atmospheric Turbulence Simulator

Phase-over-aperture model (Version 1)

Key Concept: Collapse screens and sample in the Zernike space. 20x speed up compared to split-step.

Publication:

Nicholas Chimitt and Stanley H. Chan, ‘‘Simulating Anisoplanatic Turbulence by Sampling Correlated Zernike Coefficients’’, Optical Engineering, 59(8), 083101, July 2020. (A shorter version is presented in IEEE ICCP 2020.)

Code:

MATLAB download: (URL)
Python download: (URL)
Licence: Copyright is granted for educational and research purposes. Please contact Prof Chan for licensing.

Project Page: project_turbulence_TurbSim_v1.html

Phase-to-space transform (Version 2)

Key Concept: Transform Zernike representation to PSF representation. 1000x speed up compared to split-step.

Publication:

Zhiyuan Mao, Nicholas Chimitt and Stanley H. Chan, ‘‘Accelerating Atmospheric Turbulence Simulation via Learned Phase-to-Space Transform’’, accepted to IEEE International Conference on Computer Vision, 2021.

Code:

Python download: (URL)
Licence: Copyright is granted for educational and research purposes. Please contact Prof Chan for licensing.

Project Page: project_turbulence_TurbSim_v2.html

Dense field phase-to-space transform (Version 3)

Key concept: Quantize and decouple Zernike mode and spatial mode to preserve wide sense stationarity. Enables full HD without interpolation.

Publication:

Nicholas Chimitt, Xingguang Zhang, Zhiyuan Mao, and Stanley H. Chan, ‘‘Real- Time Dense Field Phase-to-Space Simulation of Imaging through Atmospheric Turbulence’’, IEEE Trans. Computational Imaging, vol. 8, pp. 1159-1169, Dec. 2022.

Code:

Proprietary. Contact Stanley Chan for license.

Turbulence Reconstruction

Classical optimization-based approach

Key concept: Lucky imaging + blind deconvolution

Publication:

Zhiyuan Mao, Nicholas Chimitt, and Stanley H. Chan, ‘‘Image Reconstruction of Static and Dynamic Scenes through Anisoplanatic Turbulence’’, IEEE Transactions on Computational Imaging, vol. 6, pp. 1415-1428, Oct. 2020.

Code:

MATLAB download: (URL)
Licence: Copyright is granted for educational and research purposes. Please contact Prof Chan for licensing.

Project Page: project_turbulence_Recon_v1.html

Single-Frame Turbulence Mitigation

Key concept: Re-blur the reconstructed image using a turbulence simulator

Publication:

Zhiyuan Mao, Ajay Jaiswal, Zhangyang Wang, and Stanley H. Chan, ‘‘Single Frame Atmospheric Turbulence Mitigation: A Benchmark Study and A New Physics-Inspired Transformer Model’’, European Conference on Computer Vision ECCV, 2022.

Code:

Download: (URL)

Turbulence Mitigation Transformer

Key concept: Two-stage mitigation, transformer, temporal attention

Publication:

Xingguang Zhang, Zhiyuan Mao, Nicholas Chimitt, and Stanley H. Chan, ‘‘Imaging through the Atmosphere using Turbulence Mitigation Transformer’’, under review.

Theoretical Analysis

Tilt-then-Blur or Blur-then-Tilt

Key conclusion: Tilt + blur is the correct model

Publication:

Stanley H. Chan, ‘‘Tilt-then-Blur or Blur-then-Tilt? Clarifying the Atmospheric Turbulence Model’’, IEEE Signal Processing Letters.

Arbitrary Cn2 Profile

Key concept: Integrate Cn2 along the path instead of evaluating individual turbulence segments

Publication:

Nicholas Chimitt and Stanley Chan, ‘‘Anisoplanatic Optical Turbulence Simulation for Near-Continuous Cn2 Profiles without Wave Propagation’’, submitted.