VVC Subpicture RPR Cross-Layer Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current Versatile Video Coding (VVC) design faces limitations in supporting improved 360° video coding schemes, particularly in the combination of subpictures and spatial scalability, due to constraints that disallow the use of reference picture resampling (RPR) with multiple subpictures and impose restrictive alignment requirements across layers, leading to inefficiencies in coding performance and flexibility.
Innovation Solution
The proposed solution involves modifying the VVC design by introducing additional syntax elements and flags to control RPR and subpicture alignment, allowing for more flexible use of RPR across layers and enabling inter-layer prediction, while relaxing alignment constraints to permit different spatial resolutions and scaling windows within the same access unit, thereby enhancing coding efficiency and flexibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If reference picture resampling (RPR) is combined with multiple subpictures, then coding flexibility and 360° video performance are improved, but cross-layer alignment restrictions and device complexity increase
Solution Approach 1:
The patent introduces dynamic cross-layer alignment restrictions that allow different alignment behaviors for different layers. The restriction parameter can be configured to apply to all layers or selectively to output layers, enabling flexible adaptation to various coding scenarios while maintaining manageable complexity through conditional application.
Solution Approach 2:
The patent applies different alignment requirements to different layers locally. Output layers have stricter alignment restrictions to ensure proper display, while non-output layers can have relaxed restrictions. This local differentiation allows RPR with multiple subpictures to work effectively without imposing unnecessary complexity across the entire system.
2Stability of the object's composition
If cross-layer alignment restrictions are applied to all layers, then layer consistency is improved, but coding efficiency and scalability are reduced
Solution Approach 1:
The patent segments the layer set into output layers and non-output layers, applying different alignment restriction rules to each segment. Output layers must satisfy cross-layer alignment restrictions to maintain consistency for display, while non-output layers can be encoded with greater flexibility. This segmentation resolves the contradiction by localizing strict requirements only where necessary.
Solution Approach 2:
The alignment restriction is made dynamic through the parameter `cross-layer-alignment-restriction-ols-flag`, which can be configured to apply to all layers or only output layers. This dynamic configuration allows the system to adapt between different operational modes, optimizing both consistency and coding efficiency based on specific needs.
3Measurement precision
If syntax elements with non-binary values are used to specify RPR and resolution changes, then coding precision and control flexibility are improved, but syntax complexity and parsing overhead increase
Solution Approach 1:
The patent changes the parameter type from binary (0/1) to non-binary (multi-value) syntax elements for controlling RPR and resolution changes. This allows a single syntax element to encode multiple states (e.g., no resampling, horizontal resampling, vertical resampling, both), providing finer-grained control without requiring multiple separate flags. The trade-off is managed by consolidating control into fewer, more expressive parameters.
Data Source
AI summary
Several techniques for video encoding and video decoding are described. One example method includes performing a conversion between a subpicture in a video picture of a video and a bitstream of the video according to a rule. The rule specifies that, in in case a subpicture is treated as a video picture for the conversion, a cross-layer alignment restriction is applied to less than all of the multiple layers including a current layer that includes the subpicture and a subset of layers associated with the current layer.