Disparity Motion Vector Derivation for 3D Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current disparity motion vector derivation methods in 3D video coding are inefficient, particularly when dealing with multiple views, as they either incur high computational costs or do not optimize the choice of disparity motion vectors for improved coding and decoding efficiency.
Innovation Solution
A disparity motion vector derivation method that constructs lists of vectors obtained using different estimation methods and applies functions to these lists to select a final vector, enhancing the diversity and accuracy of motion vector prediction, thereby improving coding and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional disparity motion vector derivation methods are used in 3D video coding, then the coding process is simpler, but the coding efficiency and accuracy deteriorate
Solution Approach 1:
The patent segments the disparity motion vector derivation process into multiple candidate vectors obtained through different estimation methods (e.g., temporal prediction, spatial prediction, depth-map based methods). Each method generates a subset of candidate vectors, which are then evaluated and combined to select the optimal disparity motion vector. This segmentation allows the system to balance computational complexity with coding efficiency by selectively applying different estimation techniques.
Solution Approach 2:
The patent changes the parameter of estimation method diversity by introducing multiple different estimation techniques to generate candidate disparity motion vectors. Instead of relying on a single derivation method, the system varies the estimation approach (temporal, spatial, depth-based) to produce a set of candidates, from which the best vector is selected based on coding performance criteria. This parameter change improves coding efficiency while managing complexity through selective application.
2Measurement precision
If a single estimation method is used for disparity motion vector derivation, then the computational cost is lower, but the accuracy and reliability of motion vector prediction deteriorate
Solution Approach 1:
The patent applies partial action by generating multiple candidate disparity motion vectors using different estimation methods, but not all candidates are fully processed or transmitted. Instead, the system evaluates these candidates and selects only the most suitable ones for prediction, transmitting only the necessary information (e.g., index of selected candidate, residual). This partial processing approach improves prediction accuracy while controlling computational cost.
Solution Approach 2:
The patent introduces an intermediary selection mechanism that evaluates multiple candidate disparity motion vectors generated by different estimation methods. This intermediary process selects the optimal candidate based on prediction accuracy metrics, acting as a mediator between the multiple estimation methods and the final motion vector application. This approach improves accuracy by considering multiple sources while managing computational resources through selective evaluation.
3Reliability
If multiple disparity motion vectors from different estimation methods are constructed and functions are applied to select the final vector, then the accuracy and reliability of motion vector prediction improve, but the computational complexity increases
Solution Approach 1:
The patent segments the candidate disparity motion vectors into groups based on their estimation methods (temporal, spatial, depth-based). This segmentation allows the selection process to systematically evaluate candidates from different sources and combine them effectively. By organizing candidates into segments, the system improves reliability through diverse estimation while managing complexity through structured evaluation and selection.
Data Source
AI summary
A decoding method including: decoding at a current time instant a current image having at least two views respectively representative of a same scene. The decoding includes: deriving a disparity motion vector for a current block; predictively decoding the current block according to the derived disparity motion vector; and during the deriving: constructing a plurality of lists of disparity motion vectors, including at least one list in which at least two disparity motion vectors have been derived respectively according to at least two different estimation methods; applying a first function to the at least two disparity motion vectors of the at least one list, to obtain one disparity motion vector for each of the at least one list, and applying a second function to the disparity motion vectors of the plurality of lists to deliver the derived disparity motion vector.