Inter-view prediction with dynamic resolution reference pictures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current inter-view prediction techniques in multi-view video coding are limited by fixed downsampling ratios and inflexibility in specifying different resolutions for horizontal and vertical directions, which can lead to inefficiencies, especially when dealing with hierarchical B views and varying content requirements.
Innovation Solution
The proposed solution introduces dynamic resolution reference pictures and flexible downsampling ratios for horizontal and vertical directions, allowing spatial resolution to be determined based on view scalability levels, enabling more flexible inter-view prediction and improved syntax semantics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If fixed downsampling ratios are used in inter-view prediction, then the encoding process is simplified, but the encoding efficiency and adaptability to different content requirements deteriorate
Solution Approach 1:
The patent introduces dynamic resolution reference pictures where the resolution of reference pictures can be adjusted based on view scalability levels. The encoder can selectively apply different downsampling ratios (e.g., 2:1, 3:1, or no downsampling) to different views depending on their position in the scalability hierarchy and content characteristics, transforming the static fixed-ratio system into a dynamic adaptive one that optimizes encoding efficiency for each specific scenario
Solution Approach 2:
The patent changes the resolution parameter of reference pictures dynamically based on view levels. By introducing syntax elements that indicate view level position in the scalability hierarchy, the system can modify the resolution parameter adaptively - higher views may use lower resolution reference pictures while lower views use higher resolution, optimizing the balance between coding efficiency and computational complexity
2Ease of manufacture
If fixed downsampling ratios are applied to both horizontal and vertical directions, then the implementation is simpler, but the flexibility to optimize for different spatial dimensions is lost
Solution Approach 1:
The patent enables different downsampling ratios to be applied independently to horizontal and vertical directions. This allows the system to optimize resolution adaptation locally for each spatial dimension based on content characteristics - for example, applying a 2:1 downsampling horizontally while maintaining full vertical resolution, or using different ratios in different regions of the picture, thereby achieving superior adaptability compared to uniform downsampling
3Measurement precision
If all views are encoded at full resolution, then the prediction accuracy is improved, but the bit rate and storage requirements increase significantly
Solution Approach 1:
The patent introduces view level parameters that control the resolution of reference pictures used in inter-view prediction. By adjusting this parameter based on the view's position in the scalability hierarchy, the system can reduce the resolution of reference pictures for higher views, thereby decreasing bit rate and storage requirements while maintaining adequate prediction accuracy through selective application of downsampling ratios
Solution Approach 2:
The patent applies partial downsampling - not all views are downsampled, and not all directions are downsampled by the same amount. The system selectively applies downsampling only where necessary based on view level and content requirements, avoiding excessive action that would degrade prediction accuracy, while still achieving significant bit rate reduction in appropriate scenarios
4Adaptability or versatility
If different resolutions are used for different views, then the adaptability to content requirements is improved, but the complexity of the encoding and decoding process increases
Solution Approach 1:
The patent performs preliminary downsampling of reference pictures before they are used in inter-view prediction. By pre-processing the reference pictures to the appropriate resolution based on view level, the system avoids the need for complex real-time resolution adjustments during the prediction process itself, thereby reducing encoding and decoding complexity while maintaining the benefits of adaptive resolution
Data Source
AI summary
At least one disclosed method and apparatus relate to inter-view prediction with different resolution reference picture. A particular method includes accessing at least a portion of an encoded picture (614), the encoded picture being from a particular view of multiple views, and the portion being encoded at a particular resolution. The method further includes determining a particular view level based on the particular view (614), determining the particular resolution based on the particular view level (606), and decoding the portion based on the particular resolution (630).