Inter-view Motion Vector Prediction for 3D Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current multiview and 3D video coding techniques face inefficiencies in motion parameter signaling, particularly in the 3D-HEVC standard, where Motion Parameter Inheritance (MPI) assumes higher priority, leading to reduced coding efficiency due to frequent usage of other merge candidates and direct reuse of motion information from co-located blocks.
Innovation Solution
The proposed solution treats the motion information of co-located blocks in texture views as merge candidates with signaled indices, allowing them to be included anywhere in the candidate list, and uses this information as additional candidates for Advanced Motion Vector Prediction (AMVP) mode, enabling adaptive indexing and increased list sizes to improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Motion Parameter Inheritance (MPI) is assumed to have higher priority in the candidate list, then the decoding process is simplified, but coding efficiency is reduced due to frequent usage of other merge candidates
Solution Approach 1:
The patent applies dynamics by making the candidate list configuration adaptive rather than fixed. The MPI candidate is conditionally added to the merge candidate list based on availability, and its position in the list is dynamically determined rather than always prioritized. This allows the system to adapt to different coding scenarios, placing MPI candidates in positions that optimize coding efficiency while maintaining decoding simplicity when appropriate.
2Loss of substance
If motion information from co-located blocks is directly reused, then bitstream overhead is reduced, but coding precision deteriorates due to loss of motion detail
Solution Approach 1:
The patent merges motion information from multiple sources by adding MPI candidates to the merge candidate list alongside other motion vector candidates. Instead of directly reusing motion information from co-located blocks, the system combines MPI-derived motion vectors with motion vectors from spatial and temporal neighbors, allowing the decoder to select the most appropriate candidate. This merging approach reduces bitstream overhead while preserving motion detail through selective candidate selection.
Solution Approach 2:
The patent changes the parameter of motion vector prediction by introducing MPI candidates as additional options in the candidate list. Rather than using a fixed prediction method, the system modifies the available parameters by including multiple motion vector sources, allowing flexible selection based on the specific coding context. This enables the system to balance between bitstream compression and motion precision by choosing the most efficient candidate for each block.
3Productivity
If the candidate list size is increased to include more merge candidates, then coding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent segments the motion vector prediction process by separating the derivation of MPI candidates from other merge candidates. The MPI candidate is derived specifically from co-located blocks in other views, while other candidates come from spatial and temporal neighbors. This segmentation allows the system to manage a larger effective candidate pool without proportionally increasing complexity, as each segment follows a specific derivation rule. The segmented approach enables efficient candidate selection while maintaining manageable processing requirements.
Data Source
AI summary
For a depth block in a depth view component, a video coder derives a motion information candidate that comprises motion information of a corresponding texture block in a decoded texture view component, adds the motion information candidate to a candidate list for use in a motion vector prediction operation, and codes the current block based on a candidate in the candidate list.