Blind Unitary Transform Learning for Inverse Problems in Light-Field Imaging
Abstract
Light-field cameras have enabled a new class of digital post-processing techniques. Unfortunately, the sampling requirements needed to capture a 4D color light-field directly using a microlens array requires sacrificing spatial resolution and SNR in return for greater angular resolution. Because recovering the true light-field from focal-stack data is an ill-posed inverse problem, we propose using blind unitary transform learning (UTL) as a regularizer. UTL attempts to learn a set of filters that maximize the sparsity of the encoded representation. This paper investigates which dimensions of a light-field are most sparsifiable by UTL and lead to the best reconstruction performance. We apply the UTL regularizer to light-field inpainting and focal stack reconstruction problems and find it improves performance over traditional hand-crafted regularizers.
Cite
Text
Blocker and Fessler. "Blind Unitary Transform Learning for Inverse Problems in Light-Field Imaging." IEEE/CVF International Conference on Computer Vision Workshops, 2019. doi:10.1109/ICCVW.2019.00487Markdown
[Blocker and Fessler. "Blind Unitary Transform Learning for Inverse Problems in Light-Field Imaging." IEEE/CVF International Conference on Computer Vision Workshops, 2019.](https://mlanthology.org/iccvw/2019/blocker2019iccvw-blind/) doi:10.1109/ICCVW.2019.00487BibTeX
@inproceedings{blocker2019iccvw-blind,
title = {{Blind Unitary Transform Learning for Inverse Problems in Light-Field Imaging}},
author = {Blocker, Cameron J. and Fessler, Jeffrey A.},
booktitle = {IEEE/CVF International Conference on Computer Vision Workshops},
year = {2019},
pages = {3933-3942},
doi = {10.1109/ICCVW.2019.00487},
url = {https://mlanthology.org/iccvw/2019/blocker2019iccvw-blind/}
}