Geometric Partitioning Prediction With Weighted Samples in Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in reducing redundancy and computational complexity in inter prediction, particularly in constructing merge candidate lists and deriving motion parameters, which affect coding efficiency and performance.
Innovation Solution
Implementing a geometric partitioning mode (GPM) that partitions current blocks into two partitions with associated predictors, using a weighting index and equation to determine a weighting factor for sample reconstruction, and applying right shift operations to optimize the partitioning process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If geometric partitioning mode is implemented with weighting index to weighting factor conversion, then sample reconstruction precision is improved, but computational complexity increases
Solution Approach 1:
The patent applies parameter changes by introducing a weighting index parameter that is converted to weighting factors through a mathematical equation. This allows precise control over sample reconstruction by adjusting the weighting index, which directly influences the weighting factors applied to different partitions. The conversion equation transforms the index into meaningful weighting values that enhance reconstruction accuracy without requiring complex lookup tables.
Solution Approach 2:
The patent extracts the weighting factor calculation from complex lookup table operations and represents it through a simplified mathematical equation. By taking out the essential conversion relationship and expressing it algebraically, the patent reduces computational overhead while maintaining the precision needed for sample reconstruction. This extraction eliminates the need for storing and searching large weighting tables.
2Quantity of substance
If right shift operation and clipping are used for weighting factor calculation, then bandwidth and storage requirements are reduced, but calculation precision may be affected
Solution Approach 1:
The patent employs cheap computational operations (right shift and clipping) that can be performed with minimal resources. These operations are computationally inexpensive and can be executed efficiently in hardware, reducing the need for complex processing units and large memory structures. The right shift operation provides a simple way to scale weighting factors while clipping ensures they remain within valid ranges, both achieving resource efficiency.
Solution Approach 2:
The patent changes the representation and calculation method of weighting factors by using right shift operations instead of complex multiplication or division. This parameter transformation approach allows the system to work with scaled values that require less storage space and bandwidth while maintaining sufficient precision through the clipping operation that constrains values to appropriate ranges.
Data Source
AI summary
In a method for video coding in a decoder, prediction information of a current block of a current picture in a coded bitstream is decoded. The prediction information indicates a geometric partitioning mode (GPM) for the current block. The current block is partitioned into two partitions in the GPM mode, and each of the partitions is associated with a respective predictor. A weighting index for a sample of the current block is determined based on a position of the sample. A weighting factor is calculated based on the weighting index of the sample according to an equation that converts the weighting index to the weighting factor. The calculation includes performing a right shift operation based at least on a sum of the weighting index and an offset value, and clipping a result of the right shift operation to be within a predefined range. The sample is reconstructed based on the weighting factor and the predictor corresponding to the sample.