Motion Vector Candidate List Construction for 3D Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional three-dimensional video coding techniques, such as those in the High Efficiency Video Coding (HEVC) Test Model, face inefficiencies in motion vector prediction and candidate list derivation, particularly in inter-view prediction, leading to redundant processing and suboptimal performance in terms of bitrate and processing time.
Innovation Solution
The proposed method constructs motion vector or disparity vector candidate lists for advanced motion vector prediction (AMVP) and Merge mode by prioritizing candidates based on whether the target reference picture is an inter-view or non-inter-view reference picture, avoiding the use of default zero disparity vectors and optimizing candidate positions to reduce redundancy and improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional video coding techniques are applied to each single-view video sequence independently, then the coding process is simple and straightforward, but the coding performance is poor and inter-view redundancy is not exploited
Solution Approach 1:
The patent combines single-view coding with inter-view prediction by integrating disparity-compensated prediction (DCP) into the coding framework. The motion vector derivation process merges candidates from both the current view and reference views, allowing the system to exploit inter-view redundancy while maintaining a unified coding structure that balances simplicity and performance
2Reliability
If disparity-compensated prediction is used to exploit inter-view redundancy, then coding efficiency is improved, but the complexity of motion vector derivation and candidate list construction increases
Solution Approach 1:
The patent segments the motion vector candidate list into distinct categories: temporal candidates (from the same view) and inter-view candidates (from reference views). This segmentation allows the system to systematically manage the increased complexity by organizing candidates from different sources into separate groups with specific derivation rules for each group
Solution Approach 2:
The patent performs preliminary derivation of disparity vectors and motion vectors for candidate blocks before constructing the final candidate list. By pre-calculating these vectors and organizing them into candidate pools in advance, the system reduces the computational burden during the actual coding process and makes the complex derivation more manageable
3Measurement precision
If multiple candidate lists are constructed for AMVP and Merge mode, then prediction accuracy is improved, but processing time for encoding and decoding increases
Solution Approach 1:
The patent implements a selective candidate selection strategy where not all possible candidates are processed equally. The system prioritizes candidates based on their likelihood to provide good prediction, using heuristics to select a subset of promising candidates from the candidate lists. This partial action approach maintains prediction accuracy by focusing on the most relevant candidates while reducing processing time by avoiding exhaustive evaluation of all candidates
Data Source
AI summary
A method and apparatus for three-dimensional and multi-view video coding are disclosed, where the motion vector (MV) or disparity vector (DV) candidate list construction process for a block depends on whether the target reference picture corresponds to an inter-view reference picture or whether the inter-view candidate refers to an inter-view reference picture. In one embodiment, an MV or DV candidate list for a block coded in Merge mode is constructed, and an inter-view candidate in the MV or DV candidate list is set lower than the first candidate position if the inter-view candidate refers to an inter-view reference picture. In another embodiment, an MV or DV candidate list for a block coded in advanced motion vector prediction mode is constructed, and an inter-view candidate is set lower than the first candidate position if the inter-view candidate refers to an inter-view reference picture.