Video Coding Tree Partitioning With Predicted Block Cost
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face inefficiencies in determining optimal tree partitioning for video blocks, leading to suboptimal compression and increased bandwidth requirements.
Innovation Solution
A device determines the reference cost of a coding tree unit and compares it with the current block's average cost, using coefficients and motion thresholds to decide on splitting the block, thereby optimizing tree partitioning based on predicted cost.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If tree partitioning is performed without cost prediction, then coding flexibility is improved, but coding complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by calculating and storing reference costs for previously processed blocks before actual encoding. This pre-computed cost information is then used to guide partitioning decisions for current blocks, avoiding the need to evaluate all possible partitioning configurations in real-time, thus reducing coding complexity while maintaining flexibility
Solution Approach 2:
The patent changes the parameter of cost evaluation by introducing a simplified cost model that uses reference costs from previously encoded blocks. Instead of performing full rate-distortion optimization for each partitioning decision, the system uses a predictive cost parameter that can be quickly computed and compared, reducing processing complexity
2Measurement precision
If tree partitioning is performed without cost prediction, then partitioning accuracy is improved, but bandwidth and processing resources increase
Solution Approach 1:
The patent implements feedback by using the reference costs from previously processed blocks to inform partitioning decisions for current blocks. This feedback mechanism allows the encoder to make accurate partitioning decisions based on historical cost information, maintaining partitioning accuracy while reducing the computational resources and bandwidth needed for encoding
3Productivity
If reference cost calculation is performed for all blocks, then partitioning optimization is improved, but processing time increases
Solution Approach 1:
The patent extracts only the essential cost information from previously processed blocks - specifically the reference cost representing the minimum average cost per pixel. By extracting and storing only this critical parameter rather than full cost matrices or detailed partitioning information, the system achieves partitioning optimization without the processing overhead of analyzing complete cost data for every block
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
Systems, methods, and instrumentalities are disclosed for constraining performance of on tree partitioning based on predicted cost. A device (e.g., an encoder) may comprise a processor configured to perform one or more actions. The device may determine the reference cost of a reference region (e.g., a coding tree unit (CTU)) associated with a current region. The reference cost may comprise a minimum average cost per pixel of a plurality of coding units (CUs) in the reference region. The device may determine the current average cost of a current block of the current region. The current average cost may be based on a rate distortion cost of the current block and a number of pixels in the current block. The device may determine whether to evaluate a split the current block based at least on the reference cost and the current average cost of the current block. The device may encode the current block based on the determination of whether to evaluate a split the current block.