Published on:
25 April 2024
Primary Category:
Computer Vision and Pattern Recognition
Paper Authors:
Weiqi Li,
Shijie Zhao,
Bin Chen,
Xinhua Cheng,
Junlin Li,
Li Zhang,
Jian Zhang
Proposes ResVR, jointly optimizing image rescaling and VR viewport rendering for 360-degree panoramas
Develops discrete pixel sampling for irregular ERP and viewport correspondence
Introduces spherical pixel shape representation to enhance rendering
Achieves state-of-the-art viewport visual quality at low transmission overhead
Demonstrates gains over pipelines focused only on ERP upscaling
Jointly Rescaling and Rendering Panoramas for VR Viewing
This paper proposes a new framework called ResVR that jointly handles image rescaling and viewport rendering of 360-degree panoramic images to optimize visual quality when viewed through VR headsets. It develops techniques to enable end-to-end training of the full pipeline from server to headset display. A novel sampling method handles irregular correspondence between equirectangular and viewport image areas. A spherical pixel shape representation further enhances rendered view quality.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper