Motion Vector Prediction for Multi-Viewpoint Composite Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In predictive coding of images formed by combining multiple viewpoint images, motion vector prediction is inefficient due to large motion vectors at block positions across viewpoints, especially at boundaries between foreground and background, leading to increased code amounts and poor prediction performance.
Innovation Solution
A prediction device that generates and updates motion vectors by considering adjacent viewpoints, storing a first predicted motion vector and generating a second vector if conditions are met, such as the coding target block not being in the uppermost viewpoint and the viewpoints not being adjacent, to improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vector prediction is performed using conventional methods in composite images formed by combining multiple viewpoint images, then the prediction process is simple, but the prediction accuracy deteriorates due to large motion vectors at block positions across viewpoints
Solution Approach 1:
The patent segments the motion vector prediction process into multiple stages: first creating spatial predicted motion vectors from adjacent blocks, then creating temporal predicted motion vectors from reference pictures, and finally selecting the most appropriate predicted motion vector based on comparison with the actual motion vector. This segmented approach improves prediction accuracy while maintaining manageable complexity.
Solution Approach 2:
The patent introduces a new dimension to motion vector prediction by considering both spatial relationships (adjacent blocks in the same picture) and temporal relationships (blocks in reference pictures). This multi-dimensional approach allows the system to select predicted motion vectors from different sources (spatial or temporal) depending on which provides better accuracy for blocks spanning multiple viewpoints.
2Loss of information
If conventional motion vector prediction methods are used, then the coding process is simple, but the code amount increases due to poor prediction performance
Solution Approach 1:
The patent employs feedback mechanisms by comparing the actual motion vector with predicted motion vectors from both spatial and temporal directions. The system uses this comparison information to determine which predicted motion vector to use, thereby providing feedback that improves prediction accuracy and reduces the code amount needed to represent motion vector differences.
Solution Approach 2:
The patent changes the parameters of motion vector prediction by introducing viewpoint information as an additional parameter. When a block spans multiple viewpoints, the system adjusts the prediction strategy by selecting spatial predicted motion vectors from adjacent blocks within the same viewpoint, thereby adapting the prediction parameters to the specific characteristics of multi-viewpoint composite images.
3Measurement precision
If spatial predicted motion vectors are used for all blocks, then the prediction process is simple, but prediction performance deteriorates at boundaries between foreground and background
Solution Approach 1:
The patent introduces dynamics into the motion vector prediction process by making the prediction method adaptive rather than static. The system dynamically selects between spatial and temporal predicted motion vectors based on the specific characteristics of each block, particularly at boundaries between foreground and background. This dynamic selection improves prediction performance without requiring complex additional processing for each block type.
Data Source
AI summary
A prediction device for predicting a motion vector of a coding target block in coding of a composite image formed by combining multiple viewpoint images includes: a storage unit that stores a first predicted motion vector indicating a predicted motion vector candidate; and an update unit that if a predetermined condition is met, generates a second predicted motion vector that is a vector indicating a pixel in an adjacent viewpoint, the pixel being located at a position that is same as a position of the coding target block, and stores the second predicted motion vector the storage unit.