Weighted Prediction for Non-Rectangular Video Partitions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face challenges in achieving efficient compression and decompression of video data, particularly in limited bandwidth scenarios, where high video quality demands require improved compression ratios without compromising picture quality.
Innovation Solution
The method involves using weighted prediction parameters combined from fade weighting parameters and blending weighting parameters, with the values of fade weighting parameters determined by a reference index value and reference picture list, and blending weighting parameters determined by the position of a predicted sample in a predicted block, allowing for efficient encoding and decoding through the use of independent and position-dependent parameter sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If conventional video compression techniques are used, then data size is reduced, but picture quality deteriorates
Solution Approach 1:
The patent divides the prediction process into two separate parameter sets: position-independent parameters (applied uniformly across the block) and position-dependent parameters (varying across different sample positions). This segmentation allows each parameter type to be optimized independently, maintaining prediction accuracy while reducing the total number of parameters that need to be transmitted, thus preserving picture quality at lower bitrates
Solution Approach 2:
The patent applies different weighting strategies to different regions within a prediction block. Position-dependent parameters allow the prediction to adapt to local variations in the block, while position-independent parameters handle global characteristics. This local adaptation improves prediction accuracy in heterogeneous regions without requiring uniformly high compression across all areas
2Manufacturing precision
If more parameters are used for prediction, then picture quality is maintained, but encoding complexity increases
Solution Approach 1:
By segmenting the prediction parameters into position-independent and position-dependent categories, the patent reduces encoding complexity. Position-independent parameters can be derived once for the entire block without per-sample calculations, while position-dependent parameters use simple position-based formulas. This segmentation avoids the need to transmit or calculate complex per-sample parameters, maintaining picture quality while simplifying the encoding process
Solution Approach 2:
The patent changes the parameter representation by using position-based indexing for position-dependent parameters instead of transmitting actual parameter values for each sample. This parameter transformation reduces the number of bits required for encoding while maintaining the ability to reconstruct accurate predictions, thereby reducing encoding complexity without sacrificing picture quality
3Measurement precision
If position-dependent parameters are used, then prediction accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the parameter calculation into two phases: position-independent parameters are calculated once for the entire block, and position-dependent parameters are generated using simple position-based formulas during the prediction process. This segmentation avoids complex per-sample calculations while maintaining position-aware prediction accuracy, significantly reducing processing time compared to full per-sample parameter optimization
Solution Approach 2:
The patent performs preliminary calculation of position-independent parameters before the actual prediction process. These pre-calculated parameters are then reused for all samples in the block, avoiding redundant calculations. This preliminary action maintains prediction accuracy that accounts for position variations while reducing the overall processing time by eliminating repeated computational work
Data Source
AI summary
A method of an inter prediction process performed by a processing unit of a coding system comprises deriving combined linear combination parameters from two sets of linear combination parameters, where a first set of linear combination parameters is independent from the position of the to-be-predicted block and a second set of linear combination parameters depends on the position of the to-be-predicted block; and obtaining the value of the to-be-predicted sample using the combined linear combination parameters and predicted values of at least two samples belonging to different reference blocks of the to-be-predicted block.