In this talk, we introduce our novel deep learning approaches called "deep convolutional framelets" for inverse wave scattering problems originated from diffuse optical tomography, ultrasound imaging, wave scattering, etc. In particular, inspired by the recent discovery that a deep convolutional neural network is closely related to the Hankel matrix decomposition, we provide a unified deep learning approach for addressing inverse scattering problems that lead to low-rank Hankel matrix and associated deep convolutional neural networks.

This presentation is part of Minisymposium “MS24 - Data-driven approaches in imaging science (3 parts)”
organized by: Jae Kyu Choi (Institute of Natural Sciences, Shanghai Jiao Tong University) , Chenglong Bao (Yau Mathematical Sciences Center, Tsinghua University) .

Authors:

Jong Chul Ye (Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology)

Yeo Hun Yoon (KAIST)

Jaejun Yoo (KAIST )

Keywords:

deep learning, framelets, image reconstruction, inverse problems, inverse scattering, machine learning