A Deep Belief Network Approach to Scalable Compression of Light Field Data for Auto-Stereoscopic Displays


Sally Khaidem (Indian Institute of Technology, Madras), Mansi Sharma (Thapar Institute of Engineering & Technology)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

Light-field displays are a promising technology for creating immersive experiences by providing viewers with a binocular depth sensation and motion parallax. A glasses-free tensor/light field display is an area of research gaining popularity in auto-stereoscopic display technology. One approach to implementing a light field display with a good depth of field, wide viewing angles, and high resolution is to stack light attenuating layers. This approach allows for transparent layers to be used in the display, which enables users to view real-world scenes outside the display, making it ideal for augmented reality applications. This research paper proposes a compact and efficient representation of light field data using scalable compression of the binary-represented image layers, suitable for additive layered display, using a Deep Belief Network (DBN). Weighted binary images represent the optimized patterns, and the DBN further compresses the weighted binary patterns into a latent code representation, followed by encoding using H.265 codec. The experimental results validate the rate-scalable property of the proposed scheme. This process ensures that the layers can be transmitted over varying network conditions and played on various devices without compromising the viewing experience. Overall, the proposed pipeline is an effective and efficient light field encoding and compression method.

Citation

@inproceedings{Khaidem_2024_BMVC,
author    = {Sally Khaidem and Mansi Sharma},
title     = {A Deep Belief Network Approach to Scalable Compression of Light Field Data for Auto-Stereoscopic Displays},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0866.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection