Saved in:
Bibliographic Details
Main Authors: Wawerek-López, Paul, Bidgoli, Navid Mahmoudian, Frossard, Pascal, Kaup, André, Maugey, Thomas
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.13119
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912278462332928
author Wawerek-López, Paul
Bidgoli, Navid Mahmoudian
Frossard, Pascal
Kaup, André
Maugey, Thomas
author_facet Wawerek-López, Paul
Bidgoli, Navid Mahmoudian
Frossard, Pascal
Kaup, André
Maugey, Thomas
contents Developing effective 360-degree (spherical) image compression techniques is crucial for technologies like virtual reality and automated driving. This paper advances the state-of-the-art in on-the-sphere learning (OSLO) for omnidirectional image compression framework by proposing spherical attention modules, residual blocks, and a spatial autoregressive context model. These improvements achieve a 23.1% bit rate reduction in terms of WS-PSNR BD rate. Additionally, we introduce a spherical transposed convolution operator for upsampling, which reduces trainable parameters by a factor of four compared to the pixel shuffling used in the OSLO framework, while maintaining similar compression performance. Therefore, in total, our proposed method offers significant rate savings with a smaller architecture and can be applied to any spherical convolutional application.
format Preprint
id arxiv_https___arxiv_org_abs_2503_13119
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle OSLO-IC: On-the-Sphere Learned Omnidirectional Image Compression with Attention Modules and Spatial Context
Wawerek-López, Paul
Bidgoli, Navid Mahmoudian
Frossard, Pascal
Kaup, André
Maugey, Thomas
Image and Video Processing
Developing effective 360-degree (spherical) image compression techniques is crucial for technologies like virtual reality and automated driving. This paper advances the state-of-the-art in on-the-sphere learning (OSLO) for omnidirectional image compression framework by proposing spherical attention modules, residual blocks, and a spatial autoregressive context model. These improvements achieve a 23.1% bit rate reduction in terms of WS-PSNR BD rate. Additionally, we introduce a spherical transposed convolution operator for upsampling, which reduces trainable parameters by a factor of four compared to the pixel shuffling used in the OSLO framework, while maintaining similar compression performance. Therefore, in total, our proposed method offers significant rate savings with a smaller architecture and can be applied to any spherical convolutional application.
title OSLO-IC: On-the-Sphere Learned Omnidirectional Image Compression with Attention Modules and Spatial Context
topic Image and Video Processing
url https://arxiv.org/abs/2503.13119