Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures

Mary Guindy; Vamsi Kiran Adhikarla; Peter Andras Kara; Tibor Balogh; Aniko Simon

doi:10.1117/12.2618993

Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures

Mary Guindy
, Vamsi Kiran Adhikarla
, Peter Andras Kara
, Tibor Balogh
, Aniko Simon

Research output: Contribution to conference › Paper › peer-review

1 Downloads (Pure)

Abstract

High dynamic range imaging has recently become a technological trend, with numerous attempts to reconstruct HDR images and videos from low-dynamic-range data. The reconstruction of light field images is analogous to the reconstruction of HDR videos, within which consecutive frames are temporally coherent. For light field images, many similarities exist between the adjacent views, since they visualize the same scene from different angular perspectives. In this paper, we investigate the theoretical possibilities of combining CNN architectures utilized for HDR images and videos in order to enhance the outputs of HDR light field image reconstruction.

Original language	English
DOIs	https://doi.org/10.1117/12.2618993
Publication status	Published - 4 Apr 2022
Event	Big Data IV: Learning, Analytics, and Applications - Orlando, Florida, U.S. Duration: 3 Apr 2022 → 4 Apr 2022

Conference

Conference	Big Data IV: Learning, Analytics, and Applications
Period	3/04/22 → 4/04/22

Bibliographical note

Note: Published in: Proceedings Volume 12097, Big Data IV: Learning, Analytics, and Applications; (2022), ISSN 0277-786X,
ISBN 9781510650701, article no. 120970M.

Organising Body: SPIE

Keywords

Computer science and informatics

Access to Document

10.1117/12.2618993

Kara-P-51688-AAMAccepted author manuscript, 223 KBLicence: Unspecified

https://doi.org/10.1117/12.2618993

Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures
Guindy, M., Adhikarla, V. K., Kara, P. A., Balogh, T. & Simon, A., 4 Apr 2022, Published in: Proceedings Volume 12097, Big Data IV: Learning, Analytics, and Applications; (2022), ISSN 0277-786X, ISBN 9781510650701, article no. 120970M. Organising Body: SPIE Organising Body: SPIE.
Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Open Access
File
2 Downloads (Pure)

Cite this

@conference{3671f7e0789a4acaa14fe7496acae160,

title = "Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures",

abstract = "High dynamic range imaging has recently become a technological trend, with numerous attempts to reconstruct HDR images and videos from low-dynamic-range data. The reconstruction of light field images is analogous to the reconstruction of HDR videos, within which consecutive frames are temporally coherent. For light field images, many similarities exist between the adjacent views, since they visualize the same scene from different angular perspectives. In this paper, we investigate the theoretical possibilities of combining CNN architectures utilized for HDR images and videos in order to enhance the outputs of HDR light field image reconstruction.",

keywords = "Computer science and informatics",

author = "Mary Guindy and Adhikarla, \{Vamsi Kiran\} and Kara, \{Peter Andras\} and Tibor Balogh and Aniko Simon",

note = "Note: Published in: Proceedings Volume 12097, Big Data IV: Learning, Analytics, and Applications; (2022), ISSN 0277-786X, ISBN 9781510650701, article no. 120970M. Organising Body: SPIE; Big Data IV: Learning, Analytics, and Applications ; Conference date: 03-04-2022 Through 04-04-2022",

year = "2022",

month = apr,

day = "4",

doi = "10.1117/12.2618993",

language = "English",

}

TY - CONF

T1 - Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures

AU - Guindy, Mary

AU - Adhikarla, Vamsi Kiran

AU - Kara, Peter Andras

AU - Balogh, Tibor

AU - Simon, Aniko

N1 - Note: Published in: Proceedings Volume 12097, Big Data IV: Learning, Analytics, and Applications; (2022), ISSN 0277-786X, ISBN 9781510650701, article no. 120970M. Organising Body: SPIE

PY - 2022/4/4

Y1 - 2022/4/4

N2 - High dynamic range imaging has recently become a technological trend, with numerous attempts to reconstruct HDR images and videos from low-dynamic-range data. The reconstruction of light field images is analogous to the reconstruction of HDR videos, within which consecutive frames are temporally coherent. For light field images, many similarities exist between the adjacent views, since they visualize the same scene from different angular perspectives. In this paper, we investigate the theoretical possibilities of combining CNN architectures utilized for HDR images and videos in order to enhance the outputs of HDR light field image reconstruction.

AB - High dynamic range imaging has recently become a technological trend, with numerous attempts to reconstruct HDR images and videos from low-dynamic-range data. The reconstruction of light field images is analogous to the reconstruction of HDR videos, within which consecutive frames are temporally coherent. For light field images, many similarities exist between the adjacent views, since they visualize the same scene from different angular perspectives. In this paper, we investigate the theoretical possibilities of combining CNN architectures utilized for HDR images and videos in order to enhance the outputs of HDR light field image reconstruction.

KW - Computer science and informatics

U2 - 10.1117/12.2618993

DO - 10.1117/12.2618993

M3 - Paper

T2 - Big Data IV: Learning, Analytics, and Applications

Y2 - 3 April 2022 through 4 April 2022

ER -

Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures

Abstract

Conference

Bibliographical note

Keywords

Access to Document

Fingerprint

Research output

Towards reconstructing HDR light fields by combining 2D and 3D CNN architectures

Cite this