Single-Loop Decoding for Multi-View Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-view video coding requires significant decoding processing capability and large buffer sizes, leading to high complexity and power consumption, especially in mobile devices, due to the need to fully decode and store multiple views for inter-view prediction, which increases costs and resource demands.
Innovation Solution
Implementing single-loop decoding for multi-view video coding, where only the views to be displayed are fully decoded and stored, while other views are partially decoded or parsed, reducing the complexity and buffer size requirements by using constrained intra-texture prediction and signaling flags to manage the decoding process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If full decoding of all views is performed for inter-view prediction, then prediction accuracy is improved, but decoding complexity and buffer size increase significantly
Solution Approach 1:
The patent segments the decoding process into two distinct loops: a first decoding loop that processes only the current view to be displayed, and a second decoding loop that processes reference views for inter-view prediction. This segmentation allows the system to perform full decoding only where necessary (current view) while using simplified parsing for reference views, thereby maintaining prediction accuracy while reducing overall decoding complexity and buffer requirements.
Solution Approach 2:
The patent applies partial action by performing complete decoding only for the current view that needs to be displayed, while performing partial decoding (parsing only) for reference views used in inter-view prediction. This selective approach avoids the excessive action of fully decoding all views, reducing computational complexity and buffer size while still achieving adequate prediction accuracy for the displayed view.
2Reliability
If all views are stored in decoded picture buffer for inter-view prediction, then prediction reliability is improved, but buffer size requirements increase
Solution Approach 1:
The patent segments buffer management into two categories: the decoded picture buffer stores only the current view and its necessary temporal references, while a separate reference picture buffer stores parsed reference views from other views. This segmentation reduces the total buffer size required while maintaining prediction reliability, as each buffer contains only the data necessary for its specific function.
Solution Approach 2:
The patent applies partial action in buffer storage by storing fully decoded pictures only for the current view in the decoded picture buffer, while storing only parsed (partially decoded) reference pictures from other views in the reference picture buffer. This approach maintains sufficient prediction reliability while significantly reducing the quantity of data that must be stored.
3Measurement precision
If multiple views are fully decoded for inter-view prediction, then inter-view prediction quality is improved, but power consumption increases
Solution Approach 1:
The patent applies partial action by performing complete decoding operations only for the current view that requires display, while performing simplified parsing operations for reference views used in inter-view prediction. This selective decoding approach maintains adequate inter-view prediction quality while significantly reducing the computational power required, thereby lowering power consumption in mobile devices.
4Productivity
If view-first coding is used with hierarchical B GOP structure, then coding efficiency is improved, but buffer size and decoding complexity increase
Solution Approach 1:
The patent segments the decoding architecture into two independent loops that can process different GOP structures simultaneously. The first decoding loop handles the current view's GOP structure for display, while the second decoding loop handles reference views' GOP structures for prediction. This segmentation allows the system to maintain coding efficiency benefits of view-first coding with hierarchical B GOP while reducing buffer size requirements through selective processing.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A system and method for implementing low complexity multi-view video coding. According to various embodiments, single-loop decoding is applied to multi-view video coding. For N coded views, where only M of the N views are to be displayed, only those M views are required to be fully decoded and stored to a decoded picture buffer (DPB) when needed. Pictures of other views are only partially decoded or simply parsed and do not have to be stored into the DPB. Various embodiments also provide for an encoder that encodes multi-view video bitstreams in accordance with the single-loop decoding concept, as well as a decoder that utilizes single-loop decoding to decode and output on a subset of the encoded views from a multi-view bitstream.