VRCT: A Viewport Reconstruction-Based 360° Video Caching Solution for Tile-Adaptive Streaming
Ziwen Ye, Qing Li, Xiaoteng Ma, Dan Zhao, Yong Jiang, Lianbo Ma, Bo Yi, Gabriel‐Miro Muntean
Abstract
360 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{\circ}$</tex-math> </inline-formula> video streaming demands higher bandwidth and lower latency than conventional videos. Some solutions employ tile-adaptive 360 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{\circ}$</tex-math> </inline-formula> video streaming and edge caching mechanisms to improve the quality of their content delivery. However, it is hard to cache large amounts of popular content with limited cache capacity. Reconstruction technologies, which have been widely adopted for images and conventional videos, can potentially reconstruct complete tile-based viewports from partial observation for 360 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{\circ}$</tex-math> </inline-formula> videos, thus further relieving the pressure on the caching. However, it is challenging to design a flexible and efficient caching solution that supports viewport reconstruction for 360 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{\circ}$</tex-math> </inline-formula> videos. In this paper, we propose a Viewport Reconstruction-based 360 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{\circ}$</tex-math> </inline-formula> video Caching solution for Tile-adaptive streaming (VRCT). To enhance viewers’ quality of experience (QoE), a QoE-driven reconstruction trigger scheme is designed to determine whether to perform reconstruction or not based on current cache information and network conditions. To make efficient use of the cache space and facilitate the viewport reconstruction, a heuristic-based solution, named aggregation-based cache replacement scheme, is proposed to improve the probability of viewport reconstruction by carefully selecting which tiles to be stored in the given limited space. Through comprehensive experiments with a real head movement dataset, we show that the proposed VRCT increases the cache hit ratio by up to 29 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\%$</tex-math> </inline-formula> , reduces the backhaul usage by up to 44 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\%$</tex-math> </inline-formula> and improves user QoE by up to 32 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\%$</tex-math> </inline-formula> compared with other existing methods. In addition, experimental results show that our proposed cache replacement scheme facilitates viewport reconstruction and supports different video types.