ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Journal Article High-Bit-Depth Geometry Representation and Compression in MPEG Immersive Video System
Cited 1 time in scopus Download 104 time Share share facebook twitter linkedin kakaostory
Authors
Yoonseob Lee, Kwan-Jung Oh, Gwangsoon Lee, Byung Tae Oh
Issue Date
2024-12
Citation
IEEE Access, v.12, pp.189064-189072
ISSN
2169-3536
Publisher
IEEE
Language
English
Type
Journal Article
DOI
https://dx.doi.org/10.1109/ACCESS.2024.3515116
Abstract
Immersive video has been rapidly evolving as a new form of media for virtual reality (VR) applications, increasing demand for efficient compression and transmission techniques. As immersive videos comprise multiple two-dimensional videos, compressing these videos individually incurs a high computational cost. MPEG has been standardizing immersive video coding, referred to as MPEG Immersive Video (MIV), wherein the redundant parts of immersive videos between views are pruned and combined into atlas videos, which are then coded using a conventional video codec. However, the original high-bit-depth geometry information is quantized to fit the conventional video codec during this process, leading to the loss of detailed sub-bit information. This paper presents a novel high-bit-depth geometry representation method, wherein the remaining bits after quantization are assigned to the chrominance channel to improve the performance of rendering the videos. Additionally, adequate preprocessing is proposed for bit assignment in the YUV 4:2:0 format. The performed experiments indicate that the proposed method reduces the BD-rate by approximately 7% on average by improving the rendering performance without modifying the existing MIV system.
KSP Keywords
Geometry information, Geometry representation, Immersive video, Representation method, Video codec, Video system, Virtual Reality, bit depth, computational cost, two-dimensional(2D), video coding
This work is distributed under the term of Creative Commons License (CCL)
(CC BY NC ND)
CC BY NC ND