Bi-Prediction Motion Search Using Unidirectional Block Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image coding systems, such as those using the H. 264/AVC standard, require extensive arithmetic processing for motion search and coding cost calculations, especially for B slices which involve bi-prediction using two reference image blocks, leading to increased computational load without a proportional increase in compression ratio.
Innovation Solution
An image coding apparatus that includes a first prediction motion search unit for calculating coding costs based on one reference block and a second unit for using this information to determine coding costs based on two reference blocks, thereby reducing arithmetic processing while maintaining a high compression ratio by limiting bi-prediction motion search to block sizes selected from unidirectional prediction searches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If bi-prediction motion search is performed for all block sizes in B slices, then coding precision is improved, but arithmetic processing load increases enormously
Solution Approach 1:
The patent performs unidirectional prediction motion search for reference blocks A and B before bi-prediction motion search. The search center for bi-prediction is determined based on the results of unidirectional prediction, which preliminarily narrows down the search range and reduces the computational complexity of the subsequent bi-prediction motion search while maintaining coding precision
Solution Approach 2:
The patent applies different processing strategies to different block sizes. For large block sizes (16×16, 16×8, 8×16), bi-prediction motion search is performed. For small block sizes (8×8, 8×4, 4×8, 4×4), only unidirectional prediction is performed. This local differentiation reduces overall arithmetic processing load while maintaining coding precision where it matters most
2Productivity
If motion search is performed for three kinds of predictions in B slices, then coding efficiency is improved, but device complexity increases
Solution Approach 1:
The patent performs unidirectional prediction motion search for reference blocks A and B before bi-prediction motion search. The search center for bi-prediction is determined based on the results of unidirectional prediction, which preliminarily narrows down the search range and reduces the computational complexity of the subsequent bi-prediction motion search while maintaining coding efficiency
Solution Approach 2:
The patent applies different processing strategies to different block sizes. For large block sizes (16×16, 16×8, 8×16), bi-prediction motion search is performed. For small block sizes (8×8, 8×4, 4×8, 4×4), only unidirectional prediction is performed. This local differentiation reduces overall arithmetic processing load while maintaining coding efficiency where it matters most
3Measurement precision
If extensive coding/decoding processes are performed for each mode and block size, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent performs unidirectional prediction motion search for reference blocks A and B before bi-prediction motion search. The search center for bi-prediction is determined based on the results of unidirectional prediction, which preliminarily narrows down the search range and reduces the computational complexity of the subsequent bi-prediction motion search while maintaining coding precision
Solution Approach 2:
The patent applies different processing strategies to different block sizes. For large block sizes (16×16, 16×8, 8×16), bi-prediction motion search is performed. For small block sizes (8×8, 8×4, 4×8, 4×4), only unidirectional prediction is performed. This local differentiation reduces overall arithmetic processing load while maintaining coding precision where it matters most
Data Source
AI summary
An image coding apparatus and an image coding method are provided for reducing arithmetic processing while keeping a high compression ratio. A coding process necessary for the bi-prediction motion search can be effectively reduced by determining a block size for a bi-prediction motion search or a center position for a motion search on the basis of a result of a motion search of a unidirectional prediction. Further, a compression ratio can be also kept as high as possible, because the block size or the center position are obtained on the basis of the result of the motion search of the unidirectional prediction.


