Timezone: »
Inverse Problems in medical imaging and computer vision are traditionally solved using purely model-based methods. Among those variational regularization models are one of the most popular approaches. We propose a new framework for applying data-driven approaches to inverse problems, using a neural network as a regularization functional. The network learns to discriminate between the distribution of ground truth images and the distribution of unregularized reconstructions. Once trained, the network is applied to the inverse problem by solving the corresponding variational problem. Unlike other data-based approaches for inverse problems, the algorithm can be applied even if only unsupervised training data is available. Experiments demonstrate the potential of the framework for denoising on the BSDS dataset and for computer tomography reconstruction on the LIDC dataset.
Author Information
Sebastian Lunz (University of Cambridge)
Carola-Bibiane Schönlieb (Cambridge University)
Ozan Öktem (KTH - Royal Institute of Technology)
More from the Same Authors
-
2022 : Structure preserving neural networks based on ODEs »
Davide Murari · Elena Celledoni · Brynjulf Owren · Carola-Bibiane Schönlieb · Ferdia Sherry -
2022 : Self-Supervised Learning of Phenotypic Representations from Cell Images with Weak Labels »
Jan Cross-Zamirski · Guy Williams · Elizabeth Mouchet · Carola-Bibiane Schönlieb · Riku Turkki · Yinhai Wang -
2021 : Carola Schönlieb Talk »
Carola-Bibiane Schönlieb -
2021 Poster: End-to-end reconstruction meets data-driven regularization for inverse problems »
Subhadip Mukherjee · Marcello Carioni · Ozan Öktem · Carola-Bibiane Schönlieb -
2020 Poster: Deeply Learned Spectral Total Variation Decomposition »
Tamara G. Grossmann · Yury Korolev · Guy Gilboa · Carola-Bibiane Schönlieb -
2018 Poster: Banach Wasserstein GAN »
Jonas Adler · Sebastian Lunz