History-Based Affine Parameter Inheritance for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding standards face challenges in efficiently managing bandwidth due to the complexity of motion vectors in video processing, particularly with affine prediction methods that require precise calculation and storage of motion parameters.
Innovation Solution
The proposed solution involves using geometry partitions with history-based motion vector prediction (HMVP) to store and process motion vectors associated with affine models, such as 4-parameter and 6-parameter affine models, in a processor-based system, allowing for efficient buffering and derivation of affine merge candidates to improve video processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If history-based motion vector prediction (HMVP) is used to improve motion vector prediction accuracy, then video decoding quality is improved, but device complexity increases due to buffer management requirements
Solution Approach 1:
The patent segments the affine parameters into different categories (4-parameter affine model parameters and 6-parameter affine model parameters) and maintains separate buffer structures for each type. This segmentation allows for more organized buffer management and reduces the complexity of tracking and retrieving specific parameter types during video decoding.
Solution Approach 2:
The patent extracts and stores only the necessary affine parameters (a, b, c, d, e, f) from previously decoded video blocks into the HMVP buffer, rather than storing complete motion vector data structures. This extraction approach reduces buffer memory requirements and simplifies the buffer management process while maintaining prediction accuracy.
2Productivity
If affine parameters are stored in a buffer for future predictions, then prediction efficiency is improved, but memory usage increases
Solution Approach 1:
The patent applies local quality by storing different types of affine parameters (4-parameter and 6-parameter models) with their specific structural characteristics in the buffer. Each parameter type is maintained with its appropriate data structure and validation rules, allowing efficient retrieval of the specific parameter type needed for current prediction tasks while optimizing memory usage for each parameter category.
Solution Approach 2:
The patent implements a mechanism where affine parameters in the HMVP buffer are discarded when they expire or when the buffer reaches capacity, and new parameters are recovered and stored. This discarding and recovering process ensures that the buffer maintains a manageable size while continuously providing fresh prediction data for improved prediction efficiency.
3Adaptability or versatility
If multiple affine model parameters are buffered for different video blocks, then adaptability of motion prediction is improved, but difficulty of managing buffer contents increases
Solution Approach 1:
The patent introduces asymmetry in buffer management by implementing different validation and storage rules for 4-parameter affine models versus 6-parameter affine models. Each parameter type has its own buffer structure, validation criteria, and retrieval logic, which simplifies the management complexity by treating each parameter type according to its specific requirements rather than using a uniform management approach.
Solution Approach 2:
The HMVP buffer is designed with multi-functionality to handle both 4-parameter and 6-parameter affine model parameters within a single buffer structure. The buffer can store different parameter types simultaneously and provides unified access mechanisms, reducing the need for multiple separate buffers while maintaining adaptability to different video block requirements.
Data Source
AI summary
History-based affine parameters inheritance is described. In an exemplary aspect, a method for video processing includes deriving, for a conversion between a current block of video and a bitstream representation of the current block, affine related information associated with an affine model to be used by the current block based on affine related information associated with an affine model used by a first block stored in a buffer, wherein the first block is affine coded prior to the current block; and performing the conversion by using the derived affine related information.