Planar Video Prediction with Unequal Weights for Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing video coding standards, such as JVET, face inefficiencies in planar prediction due to equal weight assignment and constraints that cause deviations from an ideal flat plane, leading to suboptimal coding performance.
Innovation Solution
A method for calculating a final planar prediction using unequal weights applied to horizontal and vertical predictors, with position-dependent weighting factors and scaling factors, to improve prediction accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If equal weights are assigned to horizontal and vertical predictors in planar prediction, then the calculation is simple, but the prediction accuracy deteriorates
Solution Approach 1:
The patent applies different weights to horizontal and vertical predictors based on local characteristics. Specifically, the weight for the horizontal predictor is determined by the ratio of block width to block height, and the weight for the vertical predictor is determined by the ratio of block height to block width. This local adaptation of weights to block geometry improves prediction accuracy without significantly increasing computational complexity.
Solution Approach 2:
The patent changes the prediction parameters by introducing position-dependent weighting factors that vary according to pixel coordinates and block dimensions. The weights are calculated as functions of the block's width and height, allowing the prediction model to adapt to different block sizes and aspect ratios, thereby improving accuracy while maintaining computational efficiency.
2Device complexity
If constraints are applied to maintain a flat plane in planar prediction, then the prediction model is simple, but coding performance deteriorates
Solution Approach 1:
The patent introduces dynamic weighting factors that adapt to the specific block being predicted. Instead of using fixed equal weights, the weights are dynamically calculated based on the block's width and height ratios. This dynamic adaptation allows the prediction model to better fit the actual content characteristics, improving coding performance while maintaining reasonable model complexity.
Solution Approach 2:
The patent breaks the symmetry of equal weight assignment by introducing asymmetric weights based on block aspect ratio. When a block is wider than it is tall, the horizontal predictor receives higher weight; when taller than wide, the vertical predictor receives higher weight. This asymmetric approach better matches the natural geometry of video content and improves prediction accuracy.
3Productivity
If position-independent weights are used in planar prediction, then the calculation is efficient, but prediction accuracy for different block positions deteriorates
Solution Approach 1:
The patent implements position-dependent weighting where the weights for horizontal and vertical predictors are calculated based on the specific pixel position (x, y) within the block and the block's overall dimensions. This local adaptation allows each pixel to be predicted using weights optimized for its position, improving accuracy while the mathematical formulation maintains computational efficiency through systematic weight calculation.
Solution Approach 2:
The patent extends the weight calculation from a single scalar value to a two-dimensional function of position coordinates (x, y) within the block. By incorporating both horizontal and vertical position information along with block dimensions, the weight function adapts to the local geometry at each position, improving prediction accuracy across different block positions while maintaining calculation efficiency through a unified mathematical model.
Data Source
AI summary
A system for decoding a bitstream by a decoder that includes receiving a bitstream indicating how a coding tree unit was partitioned into coding units and identifying final coding units to be decoded represented. The system decodes the identified final coding units using a first predictor, a second predictor, and an adjustment factor. The system derives a prediction pixel value based upon the first predictor, the second predictor and together with the adjustment factor.