Picture Prediction Using Merged Motion Units for Irregular Motion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video coding/decoding solutions based on translational motion models face high computational complexity due to the assumption of uniform object motion, which does not accurately represent the diverse and irregular motion patterns in the real world, such as scaling, rotation, or parabolic motion.
Innovation Solution
A picture prediction method that selects a merged motion information unit set from candidate units using a non-translational motion model, such as affine, parabolic, or rotational models, to predict pixel values, reducing computational complexity by selecting motion information units from constraint-compliant sets of pixel samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a non-translational motion model is used to describe diverse motion patterns, then motion representation accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the picture block into multiple pixel samples (K1 pixel samples) and processes each sample independently to determine candidate motion information unit sets. This segmentation allows the complex non-translational motion model to be broken down into manageable per-pixel operations, reducing overall computational complexity while maintaining accurate motion representation for each region
Solution Approach 2:
The patent performs preliminary action by determining candidate motion information unit sets for each pixel sample before final merging. This pre-computation of candidate sets for individual pixel samples allows the system to prepare motion information in advance, reducing the complexity of the final merging and selection process while ensuring accurate non-translational motion modeling
2Device complexity
If conventional translational motion model is used, then computational complexity is reduced, but motion description accuracy deteriorates
Solution Approach 1:
The patent changes the motion model parameters from simple translational motion to non-translational motion models that include scaling, rotation, and other irregular motion patterns. By modifying the motion parameters to accommodate diverse motion types, the system achieves accurate motion description while managing computational complexity through efficient algorithms for parameter estimation and merging
3Measurement precision
If motion information from multiple pixel samples is merged, then prediction accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent merges candidate motion information unit sets from multiple pixel samples into a final motion information unit set. This merging process combines motion information from K1 different pixel samples to improve prediction accuracy by capturing diverse motion patterns across the picture block, while using systematic merging rules to control processing complexity
Solution Approach 2:
The patent uses copying by selecting motion information units from candidate sets of different pixel samples and replicating them in the merged motion information unit set. This copying approach allows efficient reuse of pre-computed candidate motion information, improving prediction accuracy through multiple sample contributions while avoiding redundant computations
Data Source
AI summary
A picture prediction method and a related apparatus are disclosed. A picture prediction method includes: determining K1 pixel samples in a picture block x, and determining a candidate motion information unit set corresponding to each pixel sample in the K1 pixel samples, where the candidate motion information unit set corresponding to each pixel sample includes at least one candidate motion information unit; determining a merged motion information unit set i including K1 motion information units, where each motion information unit in the merged motion information unit set i is selected from at least a part of motion information units in candidate motion information unit sets corresponding to different pixel samples in the K1 pixel samples; and predicting a pixel value of the picture block x by using a non-translational motion model and the merged motion information unit set i.