Home // INNOV 2019, The Eighth International Conference on Communications, Computation, Networks and Technologies // View article

A Rate-distortion Optimization Approach to Omnidirectional Video Coding for VR Systems

Authors:
Yufeng Zhou
Hua Chen
Mei Yu
Gangyi Jiang

Keywords: Omnidirectional video; rate-distortion optimization; ERP; weighted-to-spherically-uniform structure similarity

Abstract:
Virtual reality (VR) systems employ omnidirectional video to provide users with a strong sense of immersion. Compared with traditional video, omnidirectional video has the characteristics of full field of view, high resolution and immersion. However, a spherical omnidirectional video has to be projected into two-dimensional plane (e.g., common equirectangular projection (ERP) format) before encoding. This greatly limits the performance of the encoder due to the geometric distortion, content redundancy and other issues. Thus, considering the characteristics of projected omnidirectional images, an omnidirectional video coding rate-distortion optimization (RDO) method based on weighted-to-spherically-uniform structural similarity (WS-SSIM) is proposed. Specifically, according to the distortion of the internal structure similarity of the projection plane and the relationship between the spherical distortion and the projection plane distortion, the WS-SSIM is proposed to describe the distortion of the local block of the ERP image relative to the viewing sphere. Then, it is applied to the RDO process of omnidirectional video coding and adaptive selection of quantization parameters to improve vision-based coding efficiency. The experimental results show that compared with the HM16.9 test platform of HEVC standard, the proposed method can achieve significant bit rate savings under the same visual quality, which proves that the proposed method has a satisfactory effect on improving the RDO performance.

Pages: 31 to 34

Copyright: Copyright (c) IARIA, 2019

Publication date: November 24, 2019

Published in: conference

ISSN: 2326-9286

ISBN: 978-1-61208-758-0

Location: Valencia, Spain

Dates: from November 24, 2019 to November 28, 2019