Hierarchical Motion Estimation for Small-Block Vector Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion estimation methods, such as hierarchical search, face limitations in accurately estimating motion vectors for small blocks due to restricted search ranges, which can lead to suboptimal compression results in video encoding.
Innovation Solution
A method and apparatus for motion estimation using a hierarchical data structure with multiple layers, where each layer undergoes low-pass filtering and subsampling, allowing for a wide-range search on the lowest resolution layer and subsequent local searches on higher resolution layers, enabling the selection of optimal variable block modes and motion vectors with increased flexibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If hierarchical search is used to reduce computing amount and memory usage, then compression efficiency is improved, but motion vector estimation accuracy for small blocks deteriorates due to restricted search ranges
Solution Approach 1:
The patent segments the motion estimation process into multiple resolution layers (first layer with low resolution, second layer with high resolution). The first layer performs a wide-range search to establish a coarse motion vector, while the second layer performs a local search to refine the motion vector with higher precision. This segmentation allows the system to maintain both computational efficiency and accuracy by dividing the search space across different resolution levels.
Solution Approach 2:
The patent introduces a resolution dimension to the motion estimation process by creating a hierarchical structure with multiple layers of different resolutions. The first layer operates at low resolution to reduce computational complexity, while the second layer operates at high resolution to improve precision. This dimensional approach allows the system to balance between computing amount and estimation accuracy.
2Adaptability or versatility
If wide range search is performed on low resolution layer, then search flexibility is improved, but motion vectors of small blocks are still limited by small range in subsequent local search
Solution Approach 1:
The patent performs a preliminary wide-range search on the first layer to establish a coarse motion vector that defines a search window for the second layer. This preliminary action allows the system to explore a wide search space initially, then focus the refined search in the second layer around the previously identified region, maintaining both flexibility and precision.
Solution Approach 2:
The patent makes the search window dynamic by adjusting its size and position based on the motion vectors obtained from the first layer. The search window for the second layer is centered around the coarse motion vector from the first layer, allowing the system to adapt the search range dynamically based on the preliminary results, thus maintaining both flexibility and precision.
3Manufacturing precision
If original 16×16 block is split into four 8×8 smaller blocks for local search, then variable block mode accuracy is improved, but computing amount increases
Solution Approach 1:
The patent applies local quality by performing the computationally intensive block splitting and local search only in the second layer with high resolution data, while the first layer uses a simpler wide-range search. This allows the system to apply high precision where needed (in the final refinement stage) while maintaining computational efficiency in earlier stages.
Solution Approach 2:
The patent performs partial block splitting by only splitting blocks that require refined motion estimation in the second layer, rather than splitting all blocks uniformly. This selective approach reduces the overall computing complexity while still achieving high accuracy for blocks that benefit from variable block modes.
Data Source
AI summary
A method for motion estimation and the apparatus thereof are provided. The method for motion estimation uses multi-resolution hierarchial search and allows splitting the optimal block mode at the level of the lowest resolution. The method also allows further splitting of blocks during local refinement at levels of higher resolutions.