Yaqoob, Abid ORCID: 0000-0002-9541-4251 (2022) Enabling bandwidth-efficient and quality-oriented adaptive streaming for omnidirectional videos. PhD thesis, Dublin City University.
Abstract
Recently, omnidirectional (or 360°) video, Virtual Reality (VR), and multi-view videos have become wide-used streaming formats for immersive multimedia applications. The latest trends in terms of the adoption of immersive applications are due to the recent proliferation of smart displays and delivery techniques for personal
and enterprise uses, but mostly because they offer a closer-to-real-life viewing experience. Unfortunately, compared to regular monoscopic videos, 360° videos have different requirements related to content preparation, packaging, transmission, specialized viewing equipment, and display characteristics (e.g., brightness, contrast,
delay, frame rate, resolution, image quality, etc.). These aspects affect the remote transmission of such a massive amount of content and the related applications require substantial network and computational resources, which are challenging to support with the conventional transmission and rendering infrastructure. Employing state-of-the-art viewing region-based adaptive streaming techniques somehow lowers the delivery and processing requirements by performing selective transmission of certain video frame areas in response to users’ viewing information. However, they still could end up with unsatisfactory Quality of Experience (QoE) and feeble
bandwidth utilization since the accurate viewing region identification, selection, and extraction mechanisms are highly error-prone in response to highly variable viewing preferences. It is therefore essential to devise advanced streaming solutions that enable bandwidth-efficient and quality-oriented streaming over the existing best-effort networks. This work considers end-to-end video streaming in general, but focuses explicitly on designing smart adaptive transmission solutions for 360° video. First, a throughput and buffer occupancy-based HTTP Adaptive Streaming (HAS) solution is proposed to provide improved QoE for both single and multiple clients, primarily when delivered under time-varying network conditions. Secondly, a priority-aware multi-view streaming mechanism is proposed to enable differentiated quality stream-
ing with a focus on ensuring synchronous and high-quality timely playback of the multiple concurrent video streams. Next, this work covers a comprehensive survey on 360° video streaming techniques along with proposing fixed and dynamic tiling-based 360° video streaming frameworks. The proposed 360° video streaming
solutions utilize content-dependent and/or content-independent information and are empowered by fixed or flexible tiling versions selection, systematic viewport identification and selection, and intelligent bitrate adaptation mechanisms to achieve improved data transmission and QoE goals. The proposed solutions were evaluated through extensive trace-driven simulation-oriented testing against the closest state-of-the-art works. The in-depth results analysis verifies the success of proposed solutions in terms of achieving improved streaming performance.
Metadata
Item Type: | Thesis (PhD) |
---|---|
Date of Award: | November 2022 |
Refereed: | No |
Supervisor(s): | Muntean, Gabriel-Miro |
Uncontrolled Keywords: | omnidirectional video; adaptive streaming; tiling; video quality |
Subjects: | Computer Science > Multimedia systems Computer Science > Digital video Engineering > Telecommunication |
DCU Faculties and Centres: | DCU Faculties and Schools > Faculty of Engineering and Computing > School of Electronic Engineering |
Funders: | Science Foundation Ireland (SFI) Research Centres Programme under Grants 12/RC/2289 P2 (Insight Centre for Data Analytics) and 16/SP/3804 (ENABLE) |
ID Code: | 27099 |
Deposited On: | 10 Nov 2022 14:48 by Gabriel Muntean . Last Modified 10 Nov 2022 14:48 |
Documents
Full text available as:
PDF
- Archive staff only. This file is embargoed until 5 October 2025
- Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Creative Commons: Attribution-Noncommercial-No Derivative Works 4.0 93MB |
Downloads
Downloads
Downloads per month over past year
Archive Staff Only: edit this record