Low-Complexity Affine Merge Mode for VVC Memory Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in achieving high compression efficiency while managing complex motion patterns, particularly with affine motion models, which require significant memory bandwidth and computational resources.
Innovation Solution
Implementing a low complexity affine merge mode in versatile video coding (VVC) that reduces memory footprint by reusing motion data line buffers and inheriting affine motion data from neighboring blocks, allowing for efficient derivation of control point vectors without explicit signaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If affine motion models are used to handle complex motion patterns, then compression efficiency is improved, but memory bandwidth and computational resources are significantly increased
Solution Approach 1:
The patent extracts only the essential motion information needed for prediction by using merge mode, where motion vectors and control point motion vectors are inherited from neighboring blocks rather than being fully decoded. This extraction approach maintains compression efficiency while dramatically reducing memory bandwidth requirements for storing and processing motion data.
Solution Approach 2:
The motion data line buffer is designed to serve multiple purposes: storing motion vectors for both affine and non-affine modes, and supporting both merge mode and AMVP mode operations. This multi-functionality reduces the overall memory footprint required for motion data storage while maintaining full coding flexibility.
2Measurement precision
If affine motion models with explicit signaling are used, then motion prediction accuracy is improved, but device complexity and memory footprint are increased
Solution Approach 1:
In merge mode, the patent extracts motion data from neighboring blocks without requiring explicit signaling of all affine parameters. The control point motion vectors are derived by applying affine transformation to motion vectors of neighboring blocks, eliminating the need to store and process full affine parameter sets while maintaining prediction accuracy.
Solution Approach 2:
Instead of fully decoding and storing all affine motion parameters, the patent uses partial action by only inheriting necessary motion vector information from neighbors and deriving control point motion vectors through affine transformation. This partial approach achieves sufficient prediction accuracy without the full complexity of explicit affine signaling.
3Adaptability or versatility
If motion data line buffers are allocated for all modes, then coding flexibility is maintained, but memory consumption is increased
Solution Approach 1:
The motion data line buffer is designed as a universal structure that handles multiple coding modes (affine and non-affine, merge and AMVP) through a unified memory allocation strategy. The buffer stores motion vectors that can be interpreted differently based on the coding mode, eliminating the need for separate buffers for each mode while maintaining full coding flexibility.
Solution Approach 2:
The patent merges the storage requirements for different motion modes into a single unified buffer structure. By combining the storage needs for affine mode, non-affine mode, merge mode, and AMVP mode into one shared buffer, the system reduces total memory consumption while preserving adaptability to different coding scenarios.
Data Source
AI summary
In some aspects, the disclosure is directed to methods and systems for reducing memory utilization and increasing efficiency during affine merge mode for versatile video coding by utilizing motion vectors stored in a motion data line buffer for a prediction unit of a second coding tree unit neighboring a first coding tree unit to derive control point motion vectors for the first coding tree unit.