Spherical Region-wise Quality Ranking

Description

Spherical Region-wise Quality Ranking (SRQR) is a quality assessment methodology standardized in 3GPP Technical Specification 26.118 for immersive media services. It is designed specifically for omnidirectional (360-degree) video content, which is projected onto a sphere from which a user selects a viewport (a limited field of view) to watch. Unlike traditional 2D video quality metrics that evaluate the entire frame uniformly, SRQR recognizes that different regions of the spherical video have different perceptual importance depending on the user's current and potential future viewpoints.

The methodology works by first partitioning the spherical surface into a set of non-overlapping regions, typically defined by a standardized tiling scheme (e.g., based on equirectangular projection segmentation). For each of these spatial regions, a quality ranking score is computed or assigned. This ranking reflects the relative visual quality of that region compared to others in the same media presentation. The ranking can be based on objective metrics (like PSNR, SSIM applied to the region), subjective testing results, or a combination thereof. The output is a quality map or data structure that associates each spatial region with a rank.

SRQR operates in two primary modes: a content creation mode, where quality rankings are generated and embedded as metadata in the media file (e.g., in an ISOBMFF track), and a consumption mode, where a client or network element uses these rankings. During streaming, this metadata can drive viewport-adaptive streaming strategies. A streaming server or client can prioritize the delivery of high-quality data for regions currently in or predicted to enter the viewport, while potentially reducing quality for regions outside the viewport to save bandwidth. This enables efficient use of network resources while maximizing the perceived quality for the user.

Purpose & Motivation

SRQR was created to address the unique challenges of streaming immersive 360-degree video, which generates enormous amounts of data (often 4K-8K resolution or higher). Streaming the entire spherical video at uniformly high quality is prohibitively bandwidth-intensive. The key insight is that a user only views a portion (typically 90x90 degrees) at any time. Previous video quality metrics (e.g., MOS, PSNR for the whole frame) were inadequate as they did not account for this viewport-dependent perception.

The technology was motivated by the rise of Virtual Reality (VR) and augmented reality services in the 5G era, where 3GPP sought to standardize efficient delivery mechanisms. It solves the problem of how to objectively describe, compare, and optimize the quality of such content in a way that aligns with human perception. By providing a standardized region-wise ranking, it enables interoperability between content creators, network providers, and client devices for quality-aware streaming.

It addresses the limitations of non-standardized, proprietary adaptive streaming methods for VR. SRQR provides a common language for quality, facilitating features like network-assisted streaming (where the 5G core can be aware of important regions) and quality of experience (QoE) reporting. Its introduction in Release 15 was part of a broader 3GPP work item on immersive media, aiming to ensure high-quality VR/AR experiences can be delivered practically over mobile networks.

Key Features

• Defines a standardized methodology for partitioning a spherical video into regions for quality assessment
• Assigns a relative quality ranking to each spatial region of the immersive content
• Supports both objective calculation and subjective derivation of region quality ranks
• Enables the generation and storage of quality ranking metadata within media files (e.g., ISOBMFF)
• Facilitates viewport-adaptive streaming strategies by identifying high-priority regions
• Provides a basis for QoE measurement and reporting specific to 360-degree video services

Evolution Across Releases

Defining Specifications