Adaptive Weighted Template Matching for Video Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video coding, the use of non-optimal template regions during motion prediction leads to sub-optimal compression performance due to lack of similarity between neighboring template regions and the current block, resulting in inefficient bitstream compression.
Innovation Solution
The proposed solution involves adjusting the weights applied to template regions to prioritize those with higher similarity, using a weighted cost function to determine a final reference block that better matches the current block, thereby improving motion prediction accuracy and compression efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a standard template matching search is performed using neighboring blocks, then the motion prediction process is simple and fast, but the prediction accuracy deteriorates when neighboring template regions are not similar to the current block
Solution Approach 1:
The patent changes the parameter of template selection by introducing a weight value that can be adjusted based on the similarity between neighboring template regions and the current block. When similarity is high, the weight increases, allowing the template matching to rely more on neighboring blocks. When similarity is low, the weight decreases, reducing the influence of inappropriate neighboring templates. This dynamic parameter adjustment resolves the contradiction by adapting the template matching complexity to the actual content similarity, improving accuracy without always incurring high computational cost.
2Measurement precision
If template weights are adjusted to prioritize similar regions, then motion prediction accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies local quality by calculating weight values specifically for template regions based on their local similarity to the current block, rather than uniformly processing all regions. The weight calculation is performed locally for each neighboring block, and only regions with significant similarity contribute to the weighted matching. This approach improves accuracy by focusing computational effort on relevant regions while avoiding unnecessary processing of dissimilar areas, thus reducing overall processing time.
3Productivity
If all neighboring template regions are used in the matching search, then the search process is simple, but the compression performance deteriorates due to sub-optimal matching when templates are dissimilar
Solution Approach 1:
The patent introduces dynamics into the template selection process by making the template weights adaptive rather than static. The weight values are dynamically calculated based on the similarity between neighboring template regions and the current block, allowing the system to automatically adjust which templates are emphasized in the matching search. This dynamic approach improves compression efficiency by ensuring that only relevant, similar templates influence the motion prediction, while avoiding the complexity of manual template selection or fixed rigid rules.
Data Source
AI summary
In some embodiments, a method selects a first template of pixels that is outside of a first block and a second template of pixels that is outside of the first block. Then, the method selects a first region of pixels that are inside the first block and a second region of pixels that are inside the first block. A first weight is calculated based on the first template of pixels and the first region of pixels and a second weight is calculated based on the second template of pixels and the second region of pixels. The first weight and the second weight are used in a motion prediction search for a second block.