Geometric Partitioning Transition Width Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In hybrid video encoding, motion prediction and compensation based on entire blocks result in large prediction errors due to edges of moving objects not being horizontal or vertical, and existing geometric partitioning modes with fixed weight transition widths are not optimal for varying motion speeds and resolutions.
Innovation Solution
A method and apparatus that determine optimal transition width parameters for weight matrices in geometric partitioning mode by template matching, sorting candidate parameters based on error metrics to improve prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed weight transition width is used in GPM, then the device complexity is reduced, but the prediction accuracy deteriorates due to inability to adapt to varying motion speeds and resolutions
Solution Approach 1:
The patent applies the dynamics principle by transitioning from a fixed weight transition width to a variable one. The weight transition width is dynamically determined based on motion vector differences and block sizes, allowing the system to adapt to varying motion speeds and resolutions. This is achieved by calculating different transition widths for different regions and motion characteristics, thereby improving prediction accuracy without significantly increasing overall system complexity.
Solution Approach 2:
The patent implements parameter changes by modifying the weight transition width parameter based on motion characteristics. Specifically, the transition width is adjusted according to motion vector differences between blocks and their sizes. This parameter adaptation allows the weight matrix to better match the actual motion patterns in the video content, resolving the contradiction between fixed complexity and variable accuracy requirements.
2Device complexity
If motion prediction is performed on entire blocks, then the device complexity is reduced, but the prediction accuracy deteriorates due to edges not being horizontal or vertical
Solution Approach 1:
The patent applies segmentation by dividing the current block into multiple sub-blocks based on geometric partitioning. Each sub-block is then processed with appropriate weight matrices that account for local motion characteristics. This segmentation allows the system to handle non-horizontal and non-vertical edges more effectively by treating different regions separately, thereby reducing prediction errors while maintaining manageable complexity through systematic processing of divided regions.
3Measurement precision
If template matching is performed to determine optimal transition width, then the prediction accuracy is improved, but the computation time increases
Solution Approach 1:
The patent implements preliminary action by pre-defining a set of candidate weight matrices with different transition widths. Instead of performing exhaustive template matching from scratch, the system selects from these pre-prepared options based on motion characteristics. This approach reduces computation time significantly while still achieving accurate transition width selection, as the heavy computational work of creating weight matrices has already been done in advance.
Data Source
AI summary
Disclosed in the present invention are a coding method and apparatus, a decoding method and apparatus, and a coder, a decoder and a storage medium. The decoding method comprises: acquiring a geometrical partition mode (GPM) prediction parameter of the current block; determining a plurality of candidate transition width parameters according to the prediction parameter; on the basis of template matching, determining a first transition width parameter actually used by a weight matrix of the current block; obtaining weight values of pixel points of the current block according to the first transition width parameter; and performing pixel-by-pixel weighting on a first predicted value and a second predicted value of the current block according to the weight values to obtain a predicted value of the current block.