Reference Picture List for Inter-View Image Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding methods do not effectively consider inter-view prediction when encoding high-definition images, leading to suboptimal reference picture selection and reduced encoding efficiency.
Innovation Solution
Incorporating inter-view prediction into the reference picture list calculation to select more appropriate reference pictures, especially for multiview video coding, by adding reference pictures from different views to the list.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional reference picture list calculation is used without inter-view prediction, then the encoding process is simpler, but encoding efficiency is reduced
Solution Approach 1:
The patent extends the traditional reference picture list from a single-view dimension to a multi-view dimension by incorporating reference pictures from different views. This allows the encoder to search for matching blocks not only in temporal directions (past/future frames) but also in spatial directions (other camera views), effectively adding a new dimension to the reference search space and improving encoding efficiency for multiview video content.
Solution Approach 2:
The reference picture list structure is enhanced to serve multiple functions simultaneously: it maintains compatibility with traditional single-view video coding while also supporting inter-view prediction. The list can dynamically include reference pictures from both the current view and other views, making it a universal structure that adapts to different video types (single-view or multiview) without requiring separate encoding paths.
2Measurement precision
If reference pictures from different views are added to the reference picture list, then reference picture selection becomes more flexible and accurate, but the reference picture list calculation becomes more complex
Solution Approach 1:
The reference picture list is designed as a dynamic structure that adapts its composition based on the video content type. For multiview video, the list dynamically includes reference pictures from multiple views, while for single-view video, it reverts to the traditional single-view configuration. This dynamic adaptability allows the system to optimize reference picture selection accuracy for each specific scenario without permanently increasing complexity.
Solution Approach 2:
The patent applies different reference picture selection strategies to different regions or types of video content. For blocks that benefit from inter-view prediction (such as regions with similar content across views), the encoder selectively uses reference pictures from other views, while for other blocks, traditional temporal prediction is sufficient. This localized application of inter-view reference improves accuracy without uniformly increasing complexity across all blocks.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image encoding method is used for encoding a current picture on a block-by-block basis and includes: generating a reference picture list by (i) assigning a reference picture index to a reference picture referable for encoding the current picture and (ii) including the reference picture assigned the reference picture index into the reference picture list (S1001); and encoding a current block included in the current picture with reference to a reference picture that is specified, from the reference picture list, for encoding the current block (S1002), wherein, when a reference picture belonging to a reference view different from a current view to which the current picture belongs has a chance of being referenced for encoding the current picture, the reference picture belonging to the reference view is added to the reference picture list in the generating (S1001).