Abstract

The non-uniform blur of atmospheric turbulence can be modeled as a superposition of linear motion blur kernels at a patch level. We propose a regression convolutional neural network (CNN) to predict angle and length of a linear motion blur kernel for varying sized patches. We analyze the robustness of the network for different patch sizes and the performance of the network in regions where the characteristics of the blur are transitioning. Alternating patch sizes per epoch in training, we find coefficient of determination scores across a range of patch sizes of ${R^2} \gt 0.78$ for length and ${R^2} \gt 0.94$ for angle prediction. We find that blur predictions in regions overlapping two blur characteristics transition between the two characteristics as overlap changes. These results validate the use of such a network for prediction of non-uniform blur characteristics at a patch level.

© 2024 Optica Publishing Group

Learning local depth regression from defocus blur by soft-assignment encoding

Rémy Leroy, Pauline Trouvé-Peloux, Bertrand Le Saux, Benjamin Buat, and Frédéric Champagnat
Appl. Opt. 61(29) 8843-8849 (2022)

Convolutional neural network for improved event-based Shack-Hartmann wavefront reconstruction

Mitchell Grose, Jason D. Schmidt, and Keigo Hirakawa
Appl. Opt. 63(16) E35-E47 (2024)

Joint object classification and turbulence strength estimation using convolutional neural networks

Daniel A. LeMaster, Steven Leung, and Olga L. Mendoza-Schrock
Appl. Opt. 60(25) G40-G48 (2021)

Effect of patch size and network architecture on a convolutional neural network approach for automatic segmentation of OCT retinal layers

Jared Hamwood, David Alonso-Caneiro, Scott A. Read, Stephen J. Vincent, and Michael J. Collins
Biomed. Opt. Express 9(7) 3049-3066 (2018)

ProDebNet: projector deblurring using a convolutional neural network

Yuta Kageyama, Mariko Isogawa, Daisuke Iwai, and Kosuke Sato
Opt. Express 28(14) 20391-20403 (2020)